GEOthermal SEISmic AI Platform (GEOSEIS-AI) for Deep and Supercritical Geothermal Exploration

In the pursuit of deep geothermal energy (depths exceeding 3 km), the limitations of traditional surface exploration often render the subsurface "invisible." This study presents an integrated seismic exploration framework: the GEOthermal SEISmic AI Platform (GEOSEIS-AI). This platform leverages high-density microseismic monitoring networks and advanced deep-learning (DL) techniques to resolve the "four pillars" essential for geothermal development: heat sources (temperature), stress states (pressure), fracture distributions (pathways), and fluid properties. Building upon the architecture of the Real-Time Microearthquake Monitoring System (RT-MEMS) (Sun et al., 2025), GEOSEIS-AI utilizes DL phase picking and earthquake localization to accelerate the processing of massive datasets. Key seismic observables—including seismicity, focal mechanism, shear-wave splitting, and seismic tomography—are employed to directly characterize these four parameters. We demonstrate the platform's capabilities through two distinct case studies: a metamorphic region in Taiwan focusing on deep geothermal potential (Huang et al., 2023) and a volcanic region in Japan targeting supercritical energy (Tsuji et al., 2025). By mapping the spatial distribution of microearthquakes, we identify the Brittle-Ductile Transition (BDT) interface. Since seismic activity ceases as rocks transition from brittle to plastic states at high temperatures (350-400°C), the "seismic-quiet zone" serves as a proxy for the top of the heat source. Identifying these thermal upwellings is essential for targeting high-enthalpy drilling sites. By analyzing P-wave first motions with DL techniques, we resolve the local stress field and faulting styles. This provides vital data for assessing wellbore stability and distinguishing between dilated, fluid-conductive faults and compressed, sealing structures. Utilizing shear-wave splitting technique, we quantify the density and orientation of subsurface fracture networks. This provides a "pre-drilling ultrasound" that identifies high-permeability zones and informs hydraulic fracturing strategies for Enhanced Geothermal Systems (EGS). Through V_p/V_s ratio analysis derived from seismic tomography, we can differentiate between solid lithology and fluid-filled pores, and more critically, the identification of fluid phases (liquid water, steam, or melt), where low and high V_p/V_s ratios act as indicators of geothermal steam and fluids, respectively. The results show that GEOSEIS-AI significantly enhances the resolution of reservoir imaging and also provides critical insights into induced seismicity monitoring for future geothermal hydrofracturing and CO₂ injection of CCS operation.

Keywords: GEOSEIS-AI; Deep Geothermal Energy; Supercritical Energy; CCS; Deep Learning; Microseismic Monitoring; Seismicity; Focal Mechanism; Shear-Wave Splitting; V_p/V_s; Seismic Tomography.

References:

Sun, W.-F., S.-Y. Pan, Y.-H. Liu, H. Kuo-Chen, C.-S. Ku, C.-M. Lin, and C.-C.Fu  (2025). A Deep-Learning-Based Real-Time Microearthquake Monitoring System (RT-MEMS) for Taiwan. Sensors, 25(11), 3353. https://doi.org/10.3390/s25113353.

Tsuji ,T., R. Andajani, M. Kato, A. Hara, N. Aoki, S. Abe, H. Kuo-Chen, Z.-K. Guan, W.-F. Sun, S.-Y. Pan, Y.-H. Liu, K. Kitamura, J. Nishijima, and H. Inagaki  (2025) Supercritical fluid flow through permeable window and phase transitions at volcanic brittle–ductile transition zone, Commun. Earth Environ. https://doi.org/10.1038/s43247-025-02774-4.

Huang S.-Y., W.-S. Chen, L.-H. Lin, H. Kuo-Chen, C.-W. Lin, W.-H. Hsu, Y.-H. Liou (2023). Geothermal characteristics of the Paolai Hot Spring area, Taiwan. 45th New Zealand Geothermal Workshop, Auckland, New Zealand.