The Ensemble Empirical Mode Decomposition(EEMD)approach is applied to divide underground fluid observation data into high frequency component,intermediate frequency component and low frequency component. The high frequency component can be used to identify and study isolated events,such as the co-seismic response. The intermediate frequency component contains earth tide signals of diurnal tide and semidiurnal tide. The low frequency component represents trend variation characteristics of the observation in a long time. After the division,specific researches could be carried out according to frequency contents of signals,providing ideas and data foundation for noise suppression,abnormal event recognition and quantitative research of mechanisms. Based on the results,the EEMD is applied to the water level observations that were collected from 2016 to 2018 at the Xin No.10 well. The co-seismic response of 31 earthquakes with magnitude 7.0 and larger is identified from the high frequency component,and the characteristics of their maximum amplitude as a function of epicentral distance and magnitude are quantitatively analyzed.