Mining subsidence monitoring and prediction is very important for coal production safety. At present, the mainstream means of mining subsidence monitoring have certain limitations, and there is a problem that the accuracy and efficiency can not be both. UAV lidar technology can be used to quickly acquire 3D point cloud of mining area surface and establish multi-phase digital elevation model (DEM),The subsidence basin can be obtained by subtraction of two periods of DEM,It has the characteristics of high efficiency and precision. In this paper, firstly, the principle process of UAV lidar surface subsidence monitoring and the inversion method of probability integral prediction parameters are analyzed and discussed. Then, taking Tangjiahui coal mine in Inner Mongolia as an example, the flight scheme of UAV lidar is designed, the two phases of laser point cloud data are collected, and the measured data are combined, solved, fused and filtered to establish two phases of DEM, The subsidence basins in the observation period were obtained, and the dynamic parameters of the whole basin were inversed. The accuracy of DEM is 0.034mm and 0.037mm respectively; The accuracy of subsidence basin is 0.050mm. The results are relatively reliable for coal mining subsidence monitoring and prediction, and provide a case for UAV lidar application in surface monitoring.