Microseismic Scanning And Seismicity In The North Tianshan Seismic Belt

As one of the most seismically active areas,the northern Tianshan area has received extensive attention from scholars.The area has diverse tectonic environments,complex seismogenic conditions and frequent seismic events,making it an ideal place to study inland earthquakes.In recent years,several earthquakes of magnitude 5 or above have occurred in the region,resulting in certain casualties,psychological damage and economic losses.The permanent seismic network in Xinjiang is relatively sparse,and the ability to monitor earthquakes is not strong,resulting in an incomplete earthquake catalog.In recent years,the rapid development of machine learning techniques,especially deep learning,has made exciting progress in earthquake detection and phase picking applications.In this paper,the northern Tianshan Mountains(43°N ～ 46°N;80°E ～90°E)are selected as the initial research area,combined with the intensive observation of temporary seismic stations deployed in this area,the deep learning algorithm EQTransformer is used to rescan to detect seismic events.Moreover,Hypoinverse absolute positioning and hypo DD relative positioning are used to improve seismic positioning,completeness and positioning accuracy of seismic activity in the study area during 2016-2018.The study shows that within the research time window,the number of earthquakes(23,078)after re-detection and preliminary positioning in the northern Tianshan area is about three times that of the station network catalog(8,248).The new catalog covers 91.43% of the seismic events of the station network in this area.After these events were relocated by hypo DD,9268 high-precision earthquakes were detected,which showed a more concentrated earthquake distribution and could better describe the fault strike.This paper conducts a more in-depth study on the seismic activity in the Jinghe area of the north Tianshan.On the basis of deep learning detection,waveform template matching detection was also carried out in Jinghe area,and it was found that a large number of missed microseismic events could still be detected.For example,after the Jinghe Ms6.6 earthquake in 2017,22 earthquake events of magnitude 3 or higher were omitted from the network catalog.The number of final detection catalogues is 11 times that of the network catalogue,and the complete magnitude is reduced from 1.3 to 0.4,greatly improving the completeness of the catalogue.The events of catalog in the Jinghe area from 2016 to 2018 were spatially represented as three clusters,with a belt-like distribution from northeast to southwest.The two clusters from south to north are the epicenters of the 2018 Jinghe Ms5.4earthquake and the 2017 Jinghe Ms6.6 earthquake.The northernmost cluster is located in the northwest direction of the 2017 Jinghe main shock.According to the waveform characteristics,cross-correlation calculation results,event time distribution and highprecision satellite images,it is comprehensively inferred to be mining blasting.The 2017 and 2018 Jinghe earthquake location results were improved using the seismic phase data predicted by deep learning.The aftershock sequence of the 2017 Jinghe earthquake is located in the west of the main shock,about 25 km westward,and the depth is mainly between 10 and 20 km.Compared with previous research results(such as Zhai et al.,2019),the distribution of aftershock bands is clearer,and the inclination of the seismogenic fault can be quantified as the southwest 45° direction.The 2018 Jinghe earthquake detected 930 newly detected events,an increase of more than 20 times compared to the 42 aftershock events in the network catalog.The abundant aftershock location results can better characterize the 80° north dip of the fault.From the spatial distribution of earthquakes,it can be seen that the 2018 Jinghe Ms5.4earthquake was not an aftershock of the 2017 Jinghe Ms6.6 earthquake.The focal mechanisms of the two Jinghe strong earthquakes are high-angle thrusts with unilateral ruptures,but the inclinations are opposite.The static Coulomb stress change image of the 2017 Jinghe Ms6.6 earthquake shows that the seismic region of the 2018 earthquake is in the shadow area of negative stress changes,which has a certain degree of inhibition.It is speculated that there is no stress triggering correlation between the two earthquakes.The results of this paper show that the new technology of deep learning to recognize earthquakes and the waveform template matching detection method have high reliability in the study area.A relatively complete earthquake catalogue in the source area can be established by rescanning and detection.According to the distribution profile characteristics of aftershocks,the time-space migration law and seismogenic structure of the two Jinghe earthquakes are obtained.