Abstract: In order to explore the settlement laws of inner dump in open pit after construction,the subsidence coeffi- cient of inner dump is defined as the ratio of the final settlement of dump surface to the initial height of dump. The in- ner dump of Shengli No. 1 Open-pit Mine in Xilinhot City,China,is taken as the research area in this study. Forty-one high temporal resolution sentinel-1A images are used to monitor the settlement of inner dump by using the small base- line subset (SBAS) technology. On this basis,the phase space reconstruction theory in the chaos theory and the sec- ond-order Volterra adaptive filtering are introduced to realize a sin-gle-step prediction of the settlement time series. The results show that ① the subsidence profile of the inner dump is obviously semi-funnel-shaped,and the cumulative set- tlement is inversely proportional to the distance from the pit. Through the analysis of the subsidence time series,it can be concluded that the areas I and II are in the active period of subsidence,with the risk of landslide and debris flow. They are the key areas for the later subsidence monitoring. The areas III-VII enter the transition period. While the area Ⅷ has been basically stable,the area basically meets the basic requirements of land reclamation and construction of simple structures. ② By curve fitting,the subsidence coefficient of the inner dump is estimated to be 0. 639 cm / m in the observation period. ③ Verified by the maximum Lyapunov index,the four sets of settlement time series ob-tained by SBAS technology all have chaotic characteristics. ④ The phase space reconstruction theory in the chaos theory and the second-order Volterra adaptive filtering are combined to realize a single-step prediction of the settlement time se- ries. The prediction results show that it can better reflect the change trend of real value in a short time. The average ab- solute error (MAE),average relative error (MAPE) and root mean square error ( RMSE) of the first ten prediction results are all below 6% ,but with the increase of prediction steps,the prediction accuracy gradually decreases. This proves that the second-order Volterra adaptive filtering can only be used for a short-term prediction of one-dimensional settlement observation data acquired by SBAS,while the long-term prediction results are unreliable.