The occurrence of subsidence caused by mining may be a complex process that causes damage to the environment. In the last century there was significant development in prediction methods for calculating surface subsidence. However, because mining may take place by multi-seam extraction, the use of current prediction methods to obtain the distribution of subsidence is a difficult and time-consuming task. Furthermore, it is impracticable to evaluate damage accurately by this method. In this paper, a new prediction method has been developed to calculate 3D subsidence by combining a stochastic model of ground movements and a geographical information system (GIS). All the subsidence calculations are implemented by a computational program, where the components of the GIS are used to fulfil the spatial–temporal analysis function. This subsidence prediction technique has been applied to calculate ground movements resulting from 21 years of coal mining under a reservoir in Japan. Details of movement were sequentially predicted and simulated in terms of years. Furthermore, subsidence-induced damage owing to progressive horizontal strain was assessed. These values conformed to the acceptable strains in reservoir dams, thus ensuring safety against tensile failure of the concrete and consequent flooding.