Abstract: In order to explore the identification method of geomagnetic jerks and analyze their characteristics, we identified and analyzed the geomagnetic jerks in recent years based on the secular variation of geomagnetic field X, Y and Z components, which were calculated by linear fitting method based on the midnight mean of geomagnetic field from geomagnetic networks of China. The results show that the occurrence time of the reported geomagnetic jerks can be accurately identified in the analysis period by Y component, the time of the jerk occurred in 1999 may be 1998 in China, and a new jerk may occur in 2017. The secular variation of Z component is gentle, and two significant changes in secular variation took place near the year 2001 and 2013, which are early than the two big geomagnetic jerks in 2003 and 2014, respectively. This is related to the high conductivity layer of the lower mantle, which can delay the propagation of geomagnetic signals from the core-mantle boundary to the surface. The secular variation of X component presents features of many geomagnetic jerks and has a certain correlation with the secular variation of D_st index, which can be used to remove the influence component of the external space current system from the X component. Then the X component can more reliably assist the Y component in recognizing the geomagnetic jerks. However, the spatial distribution of the secular variation of the geomagnetic midnight mean show consistent variation charac-teristics with that of the geomagnetic data calculated based on the IGRF12 model on the general trend, indicating that the midnight mean of geomagnetic stations can reflect the variation characteristics of the main magnetic field in China. Moreover, the spatial distribution of secular variation of Y component in 2003 showed a significant local characteristics, which may be related to the regional characteristics of the geomagnetic jerks.