Aim: Introduced species often occupy different climates in their introduced than their native range, but to what degree do such ‘climatic niche shifts’ interfere with our ability to predict invasions? Answering this question is crucial if we are to understand the threat invasive species pose to human and natural systems, especially given the ever increasing use of species distribution models as tools for invasive species risk assessment and management. Here we investigated how strongly climatic niche shifts interfered with the transferability of native- and introduced-range species distribution models.

Location: Our dataset consisted of ~14 million occurrences distributed worldwide.

Time Period: Occurrence data were collected from online repositories dating from ca. 1600 with the vast majority being from the 20^th century. Climatic data represent means between 1970–2000.

Major Taxa Studied: Our database represented 815 terrestrial plant species.

Methods: We used ordination to identify climatic niche shifts as species moved between continents. Next, we trained separate MAXENT models using native- or introduced-range occurrences, and projected those models into each species’ introduced range. We compared the ordination and MAXENT models to determine whether niche shifts were associated with errors in MAXENT predictions.

Results: Models trained on native-range occurrences poorly predicted introduced-range occurrences, and transferability was lowest in species with large climatic niche shifts. Directional shifts in species’ predicted geographic distributions mirrored their niche dynamics. This is concerning because native-range data are often used to predict introduced-range distributions.

Main Conclusions: Our results highlight the importance of considering niche shifts when modeling the potential geographic distributions of introduced species, and cast doubt on the assumption that the climatic niche of a species can be transferred between native and invasive ranges.

Data were generated from GBIF records and Worldclim data. These were used to construct MAXENT models which were then analyzed statistically. Creation and analysis of the models follows included R scripts. Many, but not all of the data generated are included here.

Scripts can be executed sequentially. Users will need to ensure that documents are in the correct folders, following instructions given in the R scripts.