Li Ying，Lu Chang

（Institute of Earthquake Forecasting, China Earthquake Administration, Beijing 100036, China）

Abstract: Seismic fluid geochemistry serves as a sensitive probe of the coupling among active faulting, crustal deformation, and deep subsurface fluid activity. Over the past two decades, the Institute of Earthquake Forecasting, China Earthquake Administration, has established a comprehensive observation and research framework covering major active tectonic belts across China. This presentation summarizes the progressive development of this framework, from regional field surveys and long-term monitoring to intelligent, data-driven anomaly recognition. We have systematically investigated the geochemical characteristics of soil gases, fault gases, groundwater, and thermal springs, including Rn, He, CO₂, CH₄, Hg, major ions, and δ²H-δ¹⁸O isotopes. By integrating hydrochemical modelling, isotope tracing, high-pressure experiments, GPS and strain measurements, terrestrial heat-flow data, and seismic imaging, we have constrained fluid sources and migration pathways. The results distinguish meteoric water recharge, water–rock interaction, crustal radiogenic contributions, and deep crustal or mantle-derived fluid components. They further clarify how fault geometry, fracture permeability, tectonic block movement, and crustal uplift jointly regulate deep–shallow fluid coupling. Representative cases from the North China Craton, the western margin of the Ordos Block, the Qilian-Haiyuan fault zone and other structures in the northeastern Tibetan Plateau, as well as the southeastern Tibetan Plateau, demonstrate clear relationships among fluid degassing, structural segmentation, and tectonic activity. Long-term in situ observations at key wells and springs provide an essential basis for distinguishing tectonic signals from complex environmental noise. Multi-year to multi-decadal hydrological and geochemical records have been analyzed to establish site-specific background baselines and to investigate the physical mechanisms responsible for pre-, co-, and post-seismic fluid variations. Current research is advancing toward real-time, data-driven anomaly recognition. Unsupervised machine-learning methods, including clustering and outlier detection, together with Bayesian change-point analysis and multicomponent synergistic algorithms, are being applied to extract nonlinear tectonic signals from continuous hydrochemical time series and suppress false anomalies caused by meteorological and anthropogenic disturbances. Based on five years of observations at the Qujiang spring and 2.5 years of records at the Wana spring near the Xiaojiang-Red River fault junction, coordinated variations in major ions and δ²H-δ¹⁸O isotopes were identified, demonstrating the potential of adaptive multicomponent criteria for short-term and imminent earthquake forecasting. The long-term goal is to construct a physically constrained, dynamically updated, and quantitatively testable framework for short-term earthquake forecasting in China through the multidisciplinary integration of fluid geochemistry, geodesy, seismology, structural geology, and artificial intelligence. Remaining challenges include regional transferability, uncertainty quantification, and independent prospective validation.

Keywords: seismic fluid geochemistry; active faults; deep-shallow fluid coupling; continuous monitoring; anomaly detection; earthquake forecasting

Author Profile: Ying Li, Research Professor and Deputy Director of the Institute of Earthquake Forecasting, China Earthquake Administration, mainly studies fluid geochemistry in active fault zones, the origin of deep Earth fluids, and the mechanisms and applications of fluid-related earthquake precursors. Email: liying@ief.ac.cn