Investigating Spatiotemporal Variations of Subsurface Velocity Structure in the Hualien Area, Taiwan: Insights from AI-Enhanced Seismic Tomography (2012–2024)

The Hualien area, situated at the collision boundary between the Eurasian Plate and the Philippine Sea Plate, is the most seismically active region in Taiwan. Despite numerous studies, the detailed subsurface geometry and fault distribution remain incompletely resolved due to the complex tectonic interactions between the plates. This study aims to refine the 3D velocity structure using seismic data collected by the Central Weather Administration Seismological Network (CWASN) and the Taiwan Strong Motion Instrumentation Program (TSMIP) from 2012 to 2024. To handle the massive dataset and improve catalog completeness, we employed deep-learning algorithms—using EQTransformer (Mousavi et al., 2020) for phase picking and GaMMA (Zhu et al., 2022) for phase association. Subsequently, we applied the double-difference tomography method (TomoDD; Zhang and Thurber, 2003), incorporating gravity constraints to better resolve shallow velocities. We performed a sequential inversion to obtain high-resolution P- and S-wave velocity structures with a grid spacing of 5 km. Our preliminary static inversion results demonstrate high resolution in onshore regions and reveal critical structural features within the collision zone. These structural geometries are generally consistent with previous tomographic models (e.g., Huang et al., 2014), ensuring the reliability of our static velocity baseline. Building on this reliable static baseline (derived from 2012–2020 data), we further investigate temporal velocity variations (4D tomography) by integrating subsequent data from 2021–2024. By integrating the refined velocity models with relocated seismicity, we aim to provide a more detailed characterization of the complex subsurface structures and their spatiotemporal variations in this active collision zone.