This study analyzes continuous seismic waveform data from the CBS station at Changbaishan Tianchi Volcano spanning 1999 to 2023. Using seismic waveform cross-correlation and the REDPy repeating earthquake detection algorithm, it systematically examines the spatiotemporal evolution characteristics of volcanic earthquake families during two episodes of volcanic unrest: the significant unrest from 2002 to 2005 and the low-level unrest from 2020 to 2021. The results show that both unrest periods exhibited similar patterns in the number of earthquake families: stability before unrest with a background count of approximately 10, a sharp increase during unrest with 67 and 3 new families respectively, and a rapid decline after unrest. Quantitative analysis of earthquake clustering rates reveals that the clustering rate was 53% during 2002—2005 and 58% during 2020—2021. Combined with seismic source depth inversion results, both episodes of unrest exhibit typical volcanic activity characteristics, yet their source distributions differ. The earlier unrest(2002—2005)featured sources concentrated at depths of 0~5km, near the top of a shallow magma chamber, closely related to stress field adjustments induced by deep magma recharge. In contrast, the later unrest(2020—2021)showed sources extending to depths of 0~8km, with long period events concentrated at shallow depths of -2.6~2km. This indicates that the recent low-level unrest is also associated with magma system activity but likely involves more complex shallow fluid processes. This research provides new seismological criteria for assessing Changbaishan's volcanic activity and deepens understanding of the physical mechanisms behind volcanic unrest.