High-precision measurements on the temporal variations of the gravity field helps us detect the redistribution of mass in the surface and subsurface. In recent decades, with the rapid development of high-precision absolute gravimetric instruments, the temporal variation of the gravity field obtained with the Terrestrial Hybrid Repeated Gravity Observation (THRGO) system have been used more and more in tectonically active regions to investigate the crustal deformation, density change, and deep crustal fluid movement.

However, what has caused the observed gravity changes and whether these gravity changes can be used as reliable earthquake precursors are still under debate. There are many factors contributing to changes in the observational data on gravity, such as instrumental artifacts, hydrogeological effects, and topography deformations. It is challenging to extract the gravity changes that are related to the deep field source process. Therefore, dedicated data processing techniques are required because the gravity signals caused by mass transfer in deep crust are relatively weak to be observed on the surface.

In a recent study published in Communications Earth & Environments, we successfully isolate the gravity signals that are potentially related to the deep mass migration by applying the newly developed Bayesian gravity data processing methods to seven THRGO datasets before and after the M_S7.0 Lushan earthquake.

First, we use the Bayesian gravity adjustment method dedicated for the continental-scale terrestrial gravity survey, to obtain high-quality gravity data and uncertainty estimation for each station and campaign. Then, we introduce a Bayesian method to invert the apparent density of material in the source region for different time period, and find a transient increase in the gravity field about 2 years before the earthquake and a drop after the mainshock in the southern epicenter of the M_S7.0 Lushan earthquake. Such gravity changes (up to 23 μGal/yr) cannot be explained by the changes of terrestrial water storage and crustal vertical deformation (about -2 μGal/yr).

Hypothesizing that the gravity increase over the region on the south of the focal zone might be related to mass transfer in the deep crust, we use a disc-shaped equivalent source model (Figure 1) with a homogeneous density to model this transferred mass. Based on Markov Chain Monte Carlo simulations, we estimate that the equivalent source model has a radius of about 82±2 km, a thickness of about 0.7±0.1 km and a depth of about 26±5 km. This equivalent source model is also supported by many other geophysical and geochemical evidence. For example, we select the earthquakes with depths >=10 km and magnitudes M_S2.0+ from August 2010 to April 2013 (before the Lushan earthquake) in this disc-shaped region. We find that the epicenters of the selected earthquakes show a migration from the disk center to the Lushan earthquake hypocenter. The migration pattern agrees with theoretical results if we assume the underground fluid diffusion rate is about 10 m²/s. Also, such fluid diffusion is indicated by the release of ³He/⁴He and CH₄ in the southern Longmenshan fault (as shown in Figure 1). 

Fig. 1: The schematic diagram of disc-shaped source model and deep mass transfer process before the M_S7.0 Lushan earthquake. D_1-7 is the cross-section of the hypocentroid and hypocenter of the M_S7.0 Lushan earthquake. The Tibet Plateau block: TPB, the South China block: SCB.

In our apparent density model, if the density increases or decreases by 0.18‰-0.74‰ of the average crustal density, it can be detected by the terrestrial time-varying gravity observation. The pre-seismic gravity increase in the southern epicenter of the M_S7.0 Lushan earthquake is likely to be related to the large-scale deep crustal mass (fluids) transfer in a broader earthquake source area. This finding helps us understand the preparation process of great earthquakes and search for reliable seismic precursors.