Search for a command to run...
Abstract Distributions of both native and invasive species are expected to shift under future climate. Species distribution models (SDMs) are often used to explore future habitats, but sources of uncertainty including novel climate conditions may reduce the reliability of future projections. We explore the potential spread of the invasive annual grass ventenata ( Ventenata dubia ) in the western United States under both current and future climate scenarios using boosted regression tree models and 30 global climate models (GCMs). We quantify novel climate conditions, prediction variability arising from both the SDMs and GCMs, and the agreement among GCMs. Results demonstrate that currently suitable habitat is concentrated inside the invaded range of the northwest, but substantial habitat exists outside the invaded range in the Southern Rockies and southwestern US mountains. Future suitability projections vary greatly among GCMs, but GCMs commonly projected decreased suitability in the invaded range and increased suitability along higher elevations of interior mountainous areas. Climate novelty did not appear to undermine the prediction reliability in many cases where the climate–species relationship was fully represented by the occurrence data. GCM‐derived variability resulting from variation in future cool season precipitation and temperature seasonality was greatest in the Rocky Mountains. SDM‐derived variability was higher in currently suitable habitat, and few GCMs projections agreed that these areas would contain future suitable habitat. However, while prediction variability was high, many GCM projections agreed that parts of the Rocky, Wasatch, and Uinta Mountains would contain highly suitable habitat in the future. As disturbances in the interior mountains occur in coming decades, reducing some natural barriers to invasion, land managers, and conservationists will need to monitor for ventenata in post‐disturbance environments. Changes to invasion potential may not play out for several decades, but results related to current potential may have applications for early detection and rapid response planning.