Abstract: Underground coal mining leads to a failure and movement of overburden strata, moreover resulting in surface subsidence. Surface subsidence changes the land topography, land continuity and even ecological environment of the land mineed. In order to reduce such adverse impacts caused by underground mining, the second coal mine of the Zhaogu mining area is investigated as an engineering case, where deep-buried coal seam with thick alluvial layer and thin bedrock is extracted. The relation between overburden movement and surface subsidence, dynamic evolution of surface subsidence and its prediction method are thoroughly studied by using lab experiment, theoretical analysis and field measurement. The results show that the overburden strata movement of longwall face with thick alluvial layer and thin bedrock is composed of a key stratum controlled stage and a thick alluvium controlled stage. In the first stage, thin bedrock presents lamellar failure mode. Bedding separation is obvious under the key stratum while thick alluvium remains stable. In the second stage, mining induced fractures initiate in the high alluvial layer. Downward propagation leads to the full-thick rupture of the thin bedrock. Correspondingly, bedding separation vanishes in this stage. Dynamic evolving process of subsidence curves at different level of overburden are achieved. It is revealed the thick alluvium exerts strong influence on strata movement and surface subsidence. The instability of the arch structure in the alluvium results in a periodic mutation in the subsidence curve. The thin bedrock moves downward quickly after full-thick rupture. The caving rock quickly compacts densely, leads to a quick closure of the caving fractures. As a result, the bulking factor of failure materials in the caving and fracturing zones is small. Correspondingly, surface subsides quickly, leading to a large subsidence coefficient. The largest subsidence nearly reaches mining height. The subsidence area is divided into direct and indirect regions. The range of the second region is enlarged by the special overburden composition. Based on such understanding, a regional prediction method is established, which is moreover used to evaluate surface subsidence. After that, the surface subsidence basin of the target mine is monitored and reconstructed with unmanned ship, which is equipped with GPS and Sonar systems. The comparison between predicting results and monitoring data reveals the largest and average errors are about 7.0% and 2.4%,respectively. That means the method is adaptable to deep mining with thick alluvium and thin bedrock, providing a guidance for addressing subsidence problems with similar conditions.