An optimal detection method, already used for global climate change detection, is applied to regional anthropogenic climate change detection, in the particular case of France. This case was chosen for the availability of high quality, statistically homogenized near-surface temperature observations on one hand, and high-resolution climate simulations over this region on the other hand. The detection is conducted on centered patterns, in order to only focus on pattern similarities between observations and model predictions. Detection is successful for 30-year trends of summer minimal daily temperatures at a 90% confidence level, while it failed on summer maximal daily and winter temperatures. This partial success is very encouraging as it shows for the first time the detection of a climate change signal at the regional scale which resembles the one computed by a regional climate model.