GEOthermal SEISmic AI Platform (GEOSEIS-AI):AI-assisted Seismic Tomography for Geothermal Exploration in the Western Foothills of Taiwan

Geothermal exploration in tectonically active regions requires reliable imaging of subsurface structures, fracture systems, and potential heat sources. Seismic methods play a critical role in providing key constraints on buried fault geometry and geothermal-related structures.

This study applies an AI-assisted seismic workflow to seismic tomography for evaluating geothermal potential in the Western Foothills of Taiwan.Earthquake catalogs generated using AI-based detection and phase-picking algorithms were used as inputs for finite-difference travel-time tomography to construct three-dimensional P- and S-wave velocity models from the surface to 8 km depth, with an approximate spatial resolution of 1 km in the upper 6 km.

Two geothermal areas were investigated: the Tai’an area in central Taiwan and the Baolai area in southwestern Taiwan, both characterized by prominent hot spring outcroppings. A total of 63 and 49 seismic stations, respectively, recorded one month of continuous data in each area. The tomography results reveal shallow seismicity mainly distributed between 3 and 7 km depth, closely associated with mapped active faults from geological investigations. High-velocity anomalies (Vp > 5.2 km/s) observed at depths of 2–5 km are interpreted as uplifted crystalline basement or competent metamorphic rocks related to orogenic processes.

These shallow high-velocity bodies likely act as geothermal heat sources and structural controls for fluid circulation, explaining the development of surface hot springs. Our results demonstrate that AI-assisted seismic tomography provides an efficient and practical framework for geothermal exploration in complex tectonic environments.