Physical mechanisms and seismogenic processes of large continental earthquakes remain among the most challenging unresolved problems in modern geophysics. The fundamental difficulty lies in the inaccessibility of the Earth’s deep interior and the extreme complexity of fault zone processes, especially at depths of 10−15 km where most large earthquakes nucleate. Traditional earthquake genesis theories, based primarily on the elastic rebound model, have largely overlooked the role of deep underground fluids. However, accumulating evidence from induced seismicity and natural observations has demonstrated that deep fluids play a critical role in earthquake preparation, nucleation, and triggering by altering effective stress, promoting fracture propagation, and facilitating fault weakening through chemical interactions such as serpentinization and silica leaching.

This paper addresses three core scientific questions. First, how do deep fluids regulate earthquake nucleation and triggering through fault-valve mechanisms? Second, do precursory anomalies in different geophysical fields originate from a common fluid-driven process? Third, how can an operational, physics-based, multi-field integrated monitoring and prediction framework be constructed? To answer these questions, this paper systematically reviews recent advances in the study of precursory phenomena and seismogenic mechanisms of continental strong earthquakes, integrating novel perspectives from ultra-deep oil and gas exploration drilling exceeding 10 000 m, deep-seated water-rock interactions, and time-varying geophysical numerical simulations of overpressured crustal fluid migration.

The heterogeneous nature of the continental crust is fundamental to understanding strong earthquake generation. The brittle-ductile transition zone, located at depths of approximately 10−20 km, serves as a critical interface where stress is concentrated, fluids are supplied, and deformation localizes. The middle crust, composed predominantly of granitic rocks, exhibits complex rheological behavior with both brittle failure and ductile flow mechanisms. Fluids constitute approximately 3%−6% of the total crustal mass, with seismic attenuation zones confirming their presence at depths of 7−12 km. Ultra-deep drilling has provided direct evidence of brine-filled fractures and microcracks at depths of 9−12 km, and recent exploration in China has revealed fault-controlled reservoirs at depths exceeding 9 000 m.

Supercritical water exists widely in the crust, with its physical and chemical properties changing dramatically near the critical point at 374 ℃ and 22.1 MPa. The dielectric constant drops sharply from 78 at normal temperature and pressure to about 6 at the critical point, while electrical conductivity reaches a maximum. This critical singularity, when coupled with faulting, can trigger a sharp increase in thermal pressure, potentially inducing earthquakes. Water-rock interactions play a vital role in pre-seismic crustal evolution. Serpentinization produces serpentine, magnetite, and hydrogen, increasing rock volume and causing stress changes. Silica leaching in middle crust rocks can cause mineral decomposition, leading to rock weakening.

Overpressured fluids are commonly present in seismically active fault zones. Several hypotheses have been proposed for their formation, including the fluid compartments model, stress barriers model, continuous flow model, fault valves model, and fluid domains model. The fault-valve mechanism is particularly representative: tectonic compression causes fluid pressure to accumulate to the critical value for crustal failure, triggering an earthquake and fluid discharge. After coseismic stress drop, fault healing allows fluid pressure to re-accumulate, initiating a new loading cycle. This causal chain provides a key physical basis for understanding earthquake recurrence cycles and identifying precursor signals.

Regarding the difference between large and small earthquakes, using the brittle-fracture layer thickness as the criterion, small earthquakes have source scales smaller than this thickness, while large earthquakes exceed it. Deep fluid activation is proposed as a key factor in the transition from small to large ruptures. Several nucleation models are compared, including the pre-slip model, cascade-up model, rate-dependent cascade-up model, fluid-driven model migratory slow-slip model, and progressive localization model.

Strong earthquakes are typically accompanied by various geophysical anomalies. In the seismic wave field, Q value and v_P/v_S ratio anomalies can persist for months to years and extend over tens to hundreds of kilometers. In the electromagnetic field, satellite data have revealed ionospheric disturbances hours to days before major earthquakes, though challenges remain in distinguishing seismo-electromagnetic signals from non-seismic sources. In the gravity field, time-varying anomalies including gravity reversals, high-gradient zones, and four-quadrant distributions have been observed, with gravity gradient zones showing high repeatability for medium-term prediction. In the deformation field, GNSS and InSAR observations have identified precursory signals including accelerated creep and strain rate deviations.

The middle crust at depths of 15−25 km generally exhibits high-conductivity and low-velocity layers, where fluids transition from subcritical to supercritical states. The unified conclusion is that anomalies in seismic waves, electromagnetic, gravity, and deformation fields are not isolated phenomena but different manifestations of the same fluid-driven fault-valve-controlled process. The common framework of deep fluid overpressure leading to fault-valve opening, followed by crustal deformation and material migration, provides a unified explanation for all observed precursors.

Three main conclusions are drawn. First, deep fluids regulate earthquake nucleation and triggering through the fault-valve mechanism. The periodic cycling of fluid pressure between near-lithostatic and hydrostatic pressures leads to the co-evolution of fault strength, permeability, and stress state. Second, multi-geophysical field anomalies exhibit homology and staged evolution, showing a temporal sequence from long-term trend to medium-term gradient zone to short-term abrupt change. Third, different precursory parameters have different applicability scales and reliability. Gravity and GNSS deformation parameters show high repeatability and operational potential for medium-term prediction, while ULF geomagnetic and TEC anomalies remain exploratory due to strong background interference.

Research gaps include weak multi-field joint inversion models, unclear critical conditions for fluid overpressure, high false alarm rates for short-term precursors, and the need for further verification of direct correspondence between deep fluid occurrence states and geophysical anomalies. Future research should focus on constructing four-dimensional joint observation networks based on physical models, developing time-varying multi-field numerical simulation technologies, introducing statistical testing and machine learning methods, and strengthening closed-loop validation of observations, experiments, and simulations.

The key conclusion is that deep-fluid-driven multi-field coupling represents a fundamental dynamic mechanism of continental strong earthquakes. An integrated monitoring framework combining four-dimensional joint observations, physics-based modeling, and machine learning optimization is advocated to advance earthquake prediction from empirical statistics toward mechanism-driven forecasting.