To assess whether a forest protected area (PA) is effective in preventing deforestation, it is common to compare the deforestation rate inside the PA with that of its surrounding area, defined using a buffer zone of a certain distance from the PA boundary. However, this methodology is often problematic for PAs located in mountainous regions because the resulting buffer area may present very different environmental features such as topography, introducing significant bias in the final assessment. We present a new approach to address these issues, based on the assumption that the control areas that best represent a given PA in terms of similarity are located along the same mountain ridge. The proposed approach is compared with commonly used methods (fixed buffers surrounding the PA), by taking the Yunlong Tianchi Nature Reserve located in the Hengduan Mountains of China, as a case study. The results showed that elevation, slope, and topographic position index of the PA were more similar to the control area defined using our approach than those defined using a fixed buffer of 3 km and 10 km. Moreover, when performing a covariate matching analysis with the aim to pair highly similar sub-samples of the PA with those of control areas, a greater number of pairs was found with our control area than the others. These findings demonstrate how our approach is able to define a control area more comparable with its corresponding PA then traditional methods. The proposed approach is particularly suitable for those mountainous regions characterized by extreme elevational gradients hosting a high diversity of landforms, vegetation, and climatic patterns coexisting in a relatively restrained area.