Abstract: How to accurately predict earthquake is a global scientific problem that needs to be solved at present, especially for short–impending earthquake prediction, which has been the hot topic of long–term attention in the earthquake field, and obtaining effective precursor information is the key to forecast this type of earthquake. Currently, it is difficult to predict short–impending earthquake because of the following reasons: firstly, there is lack of universal precursor information which can reflect the occurrence process of strong earthquake; second, the scientific community’s understanding of the earthquake initiation mechanism is not comprehensive enough; finally, the earthquake prediction theory and its technical methods have not reached the level of application. In this context, we found that the dynamics of gravitational field at ultra–low frequencies may be highly correlated with the occurrence of short–impending earthquake. Therefore, a high–precision dynamic solid tidal gravimeter and an atmospheric tidal gravimeter were developed to capture gravity anomaly, which can detect the gravitational field changes of 1～10 µGal at ultra–low frequencies (1～30 mHz). According to the monitoring, the dynamic gravity field changes of dozens of strong earthquakes from 2010 to 2023 were successfully captured, such as the Yushu earthquake in China (7.1–magnitude), the Turkey earthquake (7.8–magnitude), and Indonesia earthquake (7.4–magnitude), etc. These data have the potential to be used as the precursor information of short–impending earthquake prediction as they can reflect the whole process of earthquake from preparation to occurrence. Furthermore, the physical mechanism of "basic stability→occluded energy storage→pre–shock calm→energy release" for strong earthquake in short–impending stage was further verified. In addition, a plausible approach is proposed to achieve the time–space–intensity prediction for the short–impending earthquake, that means a prediction model of probability, time, location, and magnitude of strong earthquake is constructed based on high–density network observation and distributed precursor information database construction, and finally realize the release of early warning information several hours before the earthquake. This study remains still in the exploration stage, but the in–depth mining of precursor information hidden by the signal of gravity field changes before the earthquake will undoubtedly bring hope to the successful prediction of short–impending strong earthquake.