Species are often observed to occur in restricted patches of particularly warm microclimate at their high latitude/altitude geographic range margin. In these areas, global warming is expected to cause small-scale expansion of the occupied area, but most previous studies of range expansion have used very coarse scale data. Using high resolution microclimate models together with detailed field surveys, we tested whether the butterfly Plebejus argus, occurring on limestone grassland in north Wales, was responding as might be expected due to climate change in the last 30-40 yr. The abundance of adult Plebejus argus at 100 m resolution in 2011 was strongly affected by elevation and near-ground temperatures in May. A statistical model including microclimate, fitted to 2011 data, was successful (67% correct) at hindcasting the occurrence of Plebejus argus in 1983 when the average May air temperature was 1.4°C cooler. However, the model was less accurate at hindcasting occurrences in 1972 (50% correct). Given the distribution of micro-sites in this landscape, we predict that further warming of approximately 1°C would make the majority of sites highly microclimatically suitable for this species. There are a growing number of long-term studies of range change, and investigations into the mechanisms driving them, but still surprisingly few that explicitly make and test predictions with independent data. Our tests are a valuable example of how accurate predictions of distribution change can be, but also of the inevitable uncertainties. Improved understanding of how well models predict will be very important to plan robust climate change adaptation measures.