Microseisms Detection,Location And Tectonic Analysis Near Jiangsu And Its Adjacent Areas

The temporal and spatial distribution of earthquakes can reveal the movement characteristics of active faults and deep structures.Precise seismic location has become one of the important means to study the active structures and buried faults.Earthquake catalogue is important database for seismicity research and seismic risk analysis.The integrity of earthquake catalogue and the accuracy of earthquake event location directly affect the reliability of subsequent studies such as seismicity analysis.In this paper,the original earthquake catalogue of Jiangsu and its adjacent areas is improved by using the match and locate method accelerated by graphics processing unit(GPU-ML).We select 211 earthquake events as template events from 447 earthquakes provided by the network center of China Seismological Bureau(hereinafter referred to as network catalogue).GPU-ML is first used to scan and detect missing earthquake events from continuous data between 2009 and2018 recorded by seismic stations in Jiangsu and its adjacent areas.A total of 1349 earthquake events are identified,which is about three times of the number of earthquake catalogue events provided by the network center The original earthquake catalogue has been improved obviously in small magnitude,and the complete magnitude of the detected earthquake catalogue has been reduced from1.9 to1.2.Then we use the double difference location method to accurately locate the detected seismic events.The results show that the range of focal depth distribution of the Jianhu earthquake swarm on the southwest side of the Hongze-Goudun fault is around 5?20km The focal depth in the northwest side of the fault is deeper than those in the southeast side,and the distribution of earthquakes show a clear linear pattern along the strike of the fault.The focal depth in Sheyang area is relatively deeper compared with those in Jianhu earthquake swarm,and focal depths are confined in a depth range within 10?25 km.meanwhile we also notice that the focal depth gradually increases from northwest to southeast.We then also use the machine learning based seismic phase recognition technology to pick up the seismic phases from one-year continuous records during 2016 in Jiangsu and its adjacent areas,and then use the rapid earthquake association and location algorithm to associate the picked seismic phases with seismic events and then locate those events.We have associate and locate2196 events,which is about 20 times of the number of seismic events given in the network catalog,and it contains more than 80%events given by the network catalogue.In addition,we compare the detected earthquake events occurred in 2016 detected by two different detection methods.The earthquake event recognition method based on machine learning detecting small earthquake events from the raw data without any prior information,can pick up more earthquake events.The GPU-ML method based on waveform crosscorrelation rely on the existing seismic templates,but the accuracy of seismic event detection is higher than that based on machine learning.