<p>To accurately monitor and locate key disaster-prone strata within overlying rock masses, a novel integrated microseismic monitoring system is developed, which combines both surface and underground components. This system addresses the spatial limitations typically encountered in conventional network deployments. However, the broader monitoring area presents new challenges, particularly in developing velocity models that can enhance the localization accuracy of mining-induced seismic events. Based on the geological borehole logging data, this paper proposes a method to reconstruct and continuously update a high-resolution regional three-dimensional P-wave velocity model that adapts to mining disturbances. First, a non-dominated sorting genetic algorithm is applied to rapidly establish the stratified structure of the initial velocity model. In this phase, geological logging curves are combined with a segmentation function that distinguishes subtle intra-strata variations from more significant inter-strata differences. Subsequently, a particle swarm optimization algorithm refines the model by incorporating the observed arrival times and locations of blasting signals. These corrections are further refined by integrating local seismic velocity tomography results, resulting in an improved model for joint underground and surface microseismic positioning. Finally, employing the time difference of arrival theory, a global particle swarm optimization algorithm is utilized to determine seismic source locations by minimizing the relative travel time differences between monitoring stations. Field tests conducted at the 2108 working face of the Shilawusu Coal Mine indicate that this approach reduces the positioning error of blasting sources by 67.32% when compared with conventional velocity models; thereby, significantly enhancing the overall accuracy of seismic event localization.</p>