Extracting weak seismic-magnetic signals from strong electromagnetic backgrounds has always been an important part of seismic-electromagnetic research. Based on the data of D/H/Z components measured through a three-axis ring-core-type fluxgate magnetometer in GLM (Geermu), JYG (Jiayuguan), and KSH (Kashi) observatories, we analyzed geomagnetic signals as precursors to some earthquakes in China using the polarization method. The most effective main frequency of seismic-electromagnetic emission was determined according to the skin effect. The results showed that only 5 out of 37 earthquakes were found to have probable signals of seismomagnetic anomalies before them. Further research revealed that the significant enhancements of polarization ratios of Y_ZH and Y_ZG (Y_ZH, the spectral power ratio of Z to H and Y_ZG, the spectral power ratio of Z to G) occurred five days prior to the Jinta earthquake (M_s 5.4) and continued for about two months. The polarization ratio of Y_ZH reached an abnormally high value about one month before some earthquakes (M_s 7.4, M_s 6.7, M_s 6.7) near the KSH station and lasted for about one week to half a month, then returned to the normal value. Similarly, some enhancements of polarization ratios of Y_ZH and Y_ZG occurred two months before the Haixi earthquake (M_s 5.0) and lasted for about one month. Analysis results showed that the enhancements prior to earthquake events may be closely related to the Jinta earthquake and a series of earthquakes near the KSH station; however, the high value of Y_ZH and Y_ZG in GLM station had nothing to do with the Haixi earthquake.

Ground-based time-variable gravimetry with high accuracy is an important approach in monitoring geodynamic processes. The uncertainty of instruments including scale factor (SF) and drift rate are the primary factors affect the quality of observation data. Differing from the conventional gravity adjustment procedure, this study adopted the modified Bayesian gravity adjustment (MBGA) method, which accounts for the nonlinear drift rate, and where the SF is considered as one of the hyperparameters estimated using Akaike’s Bayesian information criterion. Based on the terrestrial time-variable gravity datasets (2018–2020) from the southeastern Tibetan Plateau, errors caused by nonlinear drift rate and SF were processed quantitatively through analysis of the gravity difference (GD) residuals and the mutual difference of the GD. Additionally, cross validation from absolute gravity (AG) values was also applied. Results suggest that: (1) the drift rate of relavive instruments show nonlinear characteristics, and owing to their different spring features, the drift rate of CG-5 is much larger than that of LCR-G gravimeters; (2) the average bias between the original and optimized SF of the CG-5 gravimeters is approximately 169 ppm, while that of the LCR-G is no more than 63 ppm; (3) comparison of the differences in gravity values (GV) suggests that the uncertainty caused by the nonlinear drift rate is smaller than that attributable to SF. Overall, the novel approach adopted in this study was found effective in removing errors, and shown to be adaptive and robust for large-scale hybrid surface gravity campaign which providing high accuracy gravity data for the geoscience research.