Abstract: Based on georesistivity data of Dashushan and Hefei deformation station of Hefei Central Seismological Observatory, the interferential characteristics created by subway running are revealed and clarified in time domain. By analysis, it is found that when the recording track goes parallel to the subway-line, the recorded data would be most severely affected, and would be less affected when the recording track goes vertically to the subway-line. Two methods are used to eliminate the subway interference with sharp peaks and abrupt changes, one method is using the wavelet packet threshold de-noising algorithm, the other is using the average value observed at night to replace the data of 24 hour average value. The results show that the wavelet packet threshold de-noising achieves good effect for georesistivity data in Hefei deformation station, the reconstructed data can not only reduce subway influence to the least but can also reduce the background noise caused by daily interference. The method using the average value observed at night to replace the data of 24 hour average value is more appropriate for the data of Dashushan station, it can improve the quality of georesistivity observation.