AMMA-2050 is a large international multidisciplinary research project addressing the challenges of understanding how the climate of West Africa will change in future decades, with the aim of supporting climate-compatible development in the region. The project includes research on future temperature increases in the region, which will have a detrimental effect on the communities living there (e.g. on health and possibly reduced crop yields). Our work as part of the project proposes a method for refining regional temperature projections and demonstrates its application to West Africa. A focus is on characterising uncertainty more comprehensively. We find a strong, but spatially varying, inter-model correlation between regional warming in West Africa over the 21^st century and global mean warming in the Coupled Model Intercomparison Project phase 5 (CMIP5) simulations. Noting this, we calculate a transformation between a frequency distribution of CMIP5-simulated global warming values and a broader published probability distribution of global warming. The latter draws on a perturbed parameter ensemble of model simulations to account for uncertainties related to the atmosphere, ocean, carbon cycle and aerosol processes that are not well characterised by the CMIP5 ensemble. We apply the transformation to CMIP5-derived distributions for warming in different regions of West Africa to obtain regional warming distributions with longer tails than distributions estimated directly from the CMIP5 ensemble. Our results imply that CMIP5-based assessments of temperature-sensitive applications may underestimate the probability of large (and small) impacts. In some cases, the transformation that we develop for regional warming could also be applied to distributions of temperature-driven climate change impacts to ensure that they better reflect known uncertainties in future climate change.