Study On The Seismic Activity Of Haicheng Region

Seismicity research focused on the location, time, earthquake magnitude and thetendency of frequency. For the distribution of stations, velocity structure and pickinguncertainty of arrival time data, the relocation precision of earthquake is always low.In this study, we collected arrival time data recorded by the seismic network and usedthe double-difference method to relocate earthquakes and analyze location results.Based on the familiar of double-difference program and parameter settings， werelocate the1400earthquakes occurred in Haicheng area（39°N—43°N，120°E—126°E） in20years. Comparing with the original earthquake location, double-difference method has its special advantage, the root-mean-square travel-time residualis reduced to0.26s from0.74s after relocation. The results present that:1) Theearthquake distribution becomes more convergent, especially in the region between40.5°N and41°N and between122°E and123°E, and more consistent with the trendof Haicheng-Dayang river fault;2) Most of earthquakes occurred in the depth from5to24km, because there exists a low-velocity and high-conductivity zone in the crust;3) From the cross sections, many earthquakes distribute as a columnar shape along thedepth because there exist a comminuted fracture zone which extending from surfaceto25km depth.Using smoothing RI model and the BP network model to test the effectiveness offrequency prediction of seismic network between1981and2013. First usingearthquake data between1891and1991to predict the following ten years earthquakefrequency. Compare prediction with real frequency, we can find that the two modelshave good predictive ability. Especially the BP network is very suitable for processingclose to actual value problem. The data after training matches to expectations well.Errors can be reduced to a lower level through several iterations. Then we use the twomodels to predict the earthquake from2003to2013, but prediction is not very ideal.The reasons are as followed:1) RI model assumes that the earthquake frequency isconsistent, but the frequency of the earthquakes in this region is not consistent. Itmeans there are differences between the initial assumptions;2) The number of datathat can be used is limited. We find earthquake frequency in the past30years can becharacterized by interval for10years, namely10years low ten years high and tenyears low again;3) Without combining statistical forecast with the actual geologicalstructure, the prediction is just pure mathematics. In this paper, the author thinks thatthe long-term seismic observation data for earthquake scientific forecast is veryimportant. Prediction should be based on reliable observation data. Only combine itwith geological analysis can we choose appropriate prediction model for earthquakeprediction.