Abstract: The hydraulic fracturing of coal seam roofs faces a technical bottleneck associated with the precise characterization of the coupled evolution of the stress, damage, and seepage fields. To address this challenge, this study proposed a time-lapse cross-hole seismic tomography method. This method integrates high-density downhole three-component geophone arrays with dynamic velocity field reconstruction technology, establishing a dynamic monitoring paradigm characterized by "excitation, high-density acquisition, and repeated time-lapse scanning". It enables high-resolution reconstruction of physical parameters of strata under the fracturing disturbance of hard coal seam roofs. A field test was carried out in the fracturing area of the hard roof above the No. 4 coal seam in a coal mine in northern Shanxi Province. Specifically, a cross-hole array with 24-channel three-component geophones was deployed, and two repeated scans were conducted before and after fracturing. By combining the anisotropy-constrained simultaneous iterative reconstruction technique (SIRT) and the hybrid velocity modeling strategy comprising CT skeleton and attribute details, the vertical imaging resolution was improved to 5 m, and the spatial resolution was better than 2.0 m. The monitoring results show that the P-wave velocities of the fractured target interval decreased by about 100 m/s on average, with negative coherence attribute anomalies and positive root mean square amplitude anomalies identified. These results verify the mechanical responses of fracturing-induced rock breaking and fracture propagation. Cross-validation of 674 high-reliability microseismic events reveals that the deviation between the dominant strikes of the low-velocity anomalies and the microseismic event cloud was merely 2°, the volumetric intersection over union (IoU) reached up to 0.71 (95% confidence interval: 0.63~0.79), and the Hausdorff distance was determined at 18 m. These results collectively confirm the reliability of the proposed method in evaluating fracturing performance and delineating stimulated reservoir volumes (SRVs). This study overcomes the limitations inherent in existing monitoring techniques for characterization of the stress-structure-function multi-scale coupling, providing a novel technical approach for dynamic monitoring of hydraulic fracturing of hard coal mine roofs.