Many species are shifting their distributions due to climate change and to increasing international trade that allows dispersal of individuals across the globe. In the case of agricultural pests, such range shifts may heavily impact agriculture. Species distribution modelling may help to predict potential changes in pest distributions. However, these modelling strategies are subject to large uncertainties coming from different sources. Here we used the case of the tomato red spider mite (Tetranychus evansi), an invasive pest that affects some of the most important agricultural crops worldwide, to show how uncertainty may affect forecasts of the potential range of the species. We explored three aspects of uncertainty: (1) species prevalence; (2) modelling method; and (3) variability in environmental responses between mites belonging to two invasive clades of T. evansi. Consensus techniques were used to forecast the potential range of the species under current and two different climate change scenarios for 2080, and variance between model projections were mapped to identify regions of high uncertainty. We revealed large predictive variations linked to all factors, although prevalence had a greater influence than the statistical model once the best modelling strategies were selected. The major areas threatened under current conditions include tropical countries in South America and Africa, and temperate regions in North America, the Mediterranean basin and Australia. Under future scenarios, the threat shifts towards northern Europe and some other temperate regions in the Americas, whereas tropical regions in Africa present a reduced risk. Analysis of niche overlap suggests that the current differential distribution of mites of the two clades of T. evansi can be partially attributed to environmental niche differentiation. Overall this study shows how consensus strategies and analysis of niche overlap can be used jointly to draw conclusions on invasive threat considering different sources of uncertainty in species distribution modelling.