Soil temperature is a key control on belowground chemical and biological processes. Typically, models of soil temperature are developed and validated for large geographic regions. However, modeling frameworks intended for higher spatial resolutions (much finer than 1 km2) are lacking across areas of complex topography. Here we propose a simple modeling framework for predicting distributed soil temperature at high temporal (i.e., 1 h steps) and spatial (i.e., 5 × 5 m) resolutions in mountainous terrain, based on a few discrete air temperature measurements. In this context, two steps were necessary to estimate the soil temperature. First, we applied the potential temperature equation to generate the air temperature distribution from a 5 m digital elevation model and Inverse Distance Weighting interpolation. Second, we applied a hybrid model to estimate the distribution of soil temperature based on the generated air temperature surfaces. Our results show that this approach simulated the spatial distribution of soil temperature well, with a root-mean-square error ranging from ~2.1 to 2.9°C. Furthermore, our approach predicted the daily and monthly variability of soil temperature well. The proposed framework can be applied to estimate the spatial variability of soil temperature in mountainous regions where direct observations are scarce.