A novel earthquake reporting system (EarthX) was launched for real-time monitoring earthquakes at China Seismic Experimental Site (CSES) across Sichuan and Yunnan provinces, China in December 2018. The EarthX system was developed by the University of Science and Technology of China (USTC) in collaboration with the Institute of Earthquake Forecasting (IEF), China Earthquake Administration (CEA). Prof. Jie Zhang of USTC is leading the development of the advanced system, which involves multiple disciplinary teams from academia and industry. The system applies novel search engine and machine learning technologies to detect earthquakes, and determine the earthquake location, magnitude, and source focal mechanism solution within a second after receiving data. Several key ideas behind the system have been published, e.g., Zhang et al. (2014, Nature Communications), Zhang et al. (2016, Geophysics), Zhang et al. (2018, arXiv preprint arXiv:1808.09603).Different from many previous earthquake monitoring systems in operation worldwide, EarthX accesses the full waveform information of the incoming earthquake data instead of just time picks to determine the source information automatically. Fully utilizing the waveform data with the machine learning method allows extracting embedded information deep in the data and generating reliable and accurate results. Automation with the machine learning method without the use of time picks ensures the system operating 24 hours a day with consistent performance. Since the system relies on the waveform pattern rather than individual wiggles or picks in seismograms, it can handle noise, and detect and locate small events (M < 2.0) that are sometimes even difficult for data analysts to identify. In general, there should be significantly more small events than large ones in an active earthquake zone. Recent performance of the system with a few events occurring in Sichuan presents promising results and a great potential to elevate the guidance effectiveness for earthquake rescue efforts. Significant efforts for improving the system are continuing, and the system is actively upgraded.
Figure 1 shows that the EarthX system is operating in China Seismic Experimental Site with real-time data and results. Figure 2 shows a flow chart of detecting and locating earthquakes with machine learning, returning a probability of the result. The highest probability point is the event location, and the probability value tells the confidence level about the result from the system. Figure 3 explains the concept of earthquake search engine, which is similar to the image search engine on the website. Given an input seismogram, the earthquake engine is capable of finding the best match within a second from a pre-established database, and telling the source information immediately.
Figure 1: EarthX system is applied to monitor earthquakes in China Seismic Experimental Site, covering Sichuan and Yunnan Provinces. Red triangle denotes seismic stations in Yunnan, and blue triangle marks seismic stations in Sichuan.
Figure 2: The machine learning flow chart of locating earthquakes in EarthX. The input are three-component data from multiple stations, and the output is a 3D image with a probability of the event location. The highest probability point marks the event location.
Figure 3: Earthquake search engine is similar to the image search engine on the website. Earthquake search engine returns good matched results and associated source information within a second.
