Abstract: Using the measured seismic records to identify the modal parameters of high earth-rock dams is a feasible way to study the dynamic characteristics of dams. Constructing the covariance-driven stochastic subspace identification (SSI) based on the data recorded at the monitoring points of a dam during the earthquake and the stable graph by establishing multiple Hankel matrices with different dimensions can effectively eliminate false modes. The hierarchical clustering can effectively identify the natural frequency and damping ratio of the dam, and realize the automatic identification of physical modal parameters, avoiding the errors introduced by human selection. The effectiveness and accuracy of the proposed SSI method for seismic data analysis are verified by numerical examples. It is applied to the identification of modal parameters of Nuozhadu core-wall rockfill dam, and the reasonable frequency and damping ratio are obtained, indicating that the method can identify the modal parameters of the dam well and has a good engineering application, and it can provide a certain basis and reference for the finite element analysis of seismic safety of high earth-rock dams.