Abstract: Taking Ningxia regional seismic network for an example, we discussed the problems or difficulties in detecting repeating earthquakes via the waveform cross-correlation method and provided the corresponding solutions. We first created statistical models of 3D volumetric heterogeneities and then incorporated them into spectral element method to simulate 3D seismic wave propagation, so investigate the spatial distribution of the cross-correlation coefficients between earthquake pairs observed by different seismic stations, we found strong impacts of the volumetric heterogeneities and focal mechanism solution on the cross-correlation coefficients. The results show that the waveform cross-correlation method is more reliable and therefore practically useful in identify repeating earthquakes than the earthquake relocation method for the Ningxia regional seismic network. The cross-correlation coefficients are affected by the intensity of 3D heterogeneities near the hypocenter and the amplitude ratio of direct waves to coda waves. Given the same distance between two earthquakes, the stronger heterogeneities near earthquake source or stronger amplitude direct waves will result in higher waveform correlation coefficients. Thus, a higher threshold of cross-correlation coefficient is needed to reliably detecting repeating earthquakes. In summary, based on the data of Ningxia regional seismic network with an average station spacing of 30−50 km, choosing a proper threshold value of cross-correlation coefficient by only taking the time window of coda waves or data with weak direct P-waves, waveform cross-correlation method could reliably identify repeating earthquakes with a magnitude range ofM_L1.0−3.0, thus providing basis for the surveillance of repeating earthquakes.