An Ms7.0 earthquake struck Jiuzhaigou (China) on 8 August 2017. The epicenter was in the eastern margin of the Tibetan Plateau, an area covered by a dense time-varying gravity observation network. Data from seven repeated high-precision hybrid gravity surveys (2014–2017) allowed the microGal-level time-varying gravity signal to be obtained at a resolution better than 75 km using the modified Bayesian gravity adjustment method. The “equivalent source” model inversion method in spherical coordinates was adopted to obtain the near-crust apparent density variations before the earthquake. A major gravity change occurred from the southwest to the northeast of the eastern Tibetan Plateau approximately 2 years before the earthquake, and a substantial gravity gradient zone was consistent with the tectonic trend that gradually appeared within the focal area of the Jiuzhaigou earthquake during 2015–2016. Factors that might cause such regional gravitational changes (e.g., vertical crustal deformation and variations in near-surface water distributions) were studied. The results suggest that gravity effects contributed by these known factors were insufficient to produce gravity changes as big as those observed, which might be related to the process of fluid material redistribution in the crust. Regional change of the gravity field has precursory significance for high-risk earthquake areas and it could be used as a candidate precursor for annual medium-term earthquake prediction.

With the advancement of high-precision absolute gravimeters, absolute/relative hybrid observation systems now enable the detection of terrestrial time-varying gravity signals, offering insights into density changes and mass redistribution within seismogenic zones. However, extracting weak gravity signals related to earthquakes remains challenging due to complex surface processes and unclear mechanisms. This study investigates gravity changes before three major earthquakes, such as the 2013 Lushan Ms7.0, 2021 Yangbi Ms6.4, and 2022 Menyuan Ms6.9 events on the eastern Tibetan Plateau. To address issues such as nonlinear drift and scale-factor variability in relative gravimeters, a Bayesian gravity adjustment method is introduced to enable quantification of observational uncertainties. Then, after correcting for hydrological and vertical deformation effects, residual gravity changes possible related to deep tectonic mass transport are derived. For quantitative description, a mass source model inversion method is used to quantify potential links between gravity changes and crustal mass transfer in seismogenic regions. Finally, combined with historical seismicity, velocity, and electrical structure models, we constructed several mass source models for understanding the mechanism of deep mass migration before strong earthquakes.