Description : We proved theoretically that geodetic height, measured with Global Positioning System (GPS), can be applied directly to monitor coal mine subsidence. Based on a Support Vector Machine (SVM) model, we built a regional geoid model with a Gaussian Radial Basis Function (RBF) and the technical scheme for GPS coal mine subsidence monitoring is presented to provide subsidence information for updating the regional Digital Elevation Model (DEM). The theory proposed was applied to monitor mining subsidence in an Inner Mongolia coal mine in China. The scheme established an accurate GPS reference network and a comprehensive leveling conjunction provided the normal height of all GPS control points. According to the case study, the SVM model to establish geoid-model is better than a polynomial fit or a Genetic Algorithm based Back Propagation (GA-BP) neural network. GPS-RTK measurements of coal mine subsidence information can be quickly acquired for updating the DEM.