Abstract: Using 9 284 Pg phase arrival times of 731 earthquakes occurred in the eastern margin of the Tibetan Plateau recorded by Sichuan and Gansu Seismic Networks from August 1, 2017 to December 31, 2017, this paper determine “the minimum one-dimensional velocity model” for the area by inversion. “The minimum one-dimensional velocity model” and the selected comparative velocity model are used to establish the comparative models. Taking the Jiuzhaigou earthquake sequence as the research object, we quantitatively discuss the effects of two velocity models on the positioning results in absolute seismic location and relative seismic location methods. The results show that “the minimum one-dimensional velocity model” obtained by inversion can effectively reduce the root-mean-square residuals of travel time in relocation. Absolute seismic location method is more dependent on one-dimensional velocity model than relative seismic location. Furthermore, one-dimensional velocity model will directly affect the distribution pattern of hypocenters in absolute location, but only adjust the relative position of seismic events in relative location. In absolute location of earthquakes, the earthquake with larger magnitude is more sensitive to velocity model, which is not obvious in relative location. Therefore, the combination of absolute location method and relative location method is the best strategy in earthquake relocation.