This study investigated the importance of the use of appropriate species distribution records in projecting potential distributions under climate change using comparative bioclimatic models and alternative sets of data (native and exotic) to project a species in a new environment. We built bioclimatic models for date palm ('Phoenix dactylifera' L.), using the MaxEnt correlative model and the CLIMEX mechanistic niche model, and fitted the models using three training data sets: native data only, exotic data only and entire data. We compared the ability of the different data sets using the different modelling approaches to project suitable climate envelope for independent records of the species at a global scale. We found that the output of projected species distributions was closely related to the modelling approach as well as the specific categorized distribution of species data used (native data only, exotic data only and entire data).