Real-Time Earthquake Location Based on the Kalman Filter Formulation

Seismic location is an essential task for earthquake monitoring. The general practice is to locate earthquakes using arrival times from all recorded stations. However, this is not well suited for real-time applications such as the earthquake early warning, where earthquake locations need to be determined and updated as more stations are triggered. Here we have developed a real-time linear location method based on a Kalman filter formulation. It updates location and its uncertainty whenever a station is triggered. We have demonstrated its effectiveness with synthetic and real data sets in Parkfield, California in a retrospective mode. The tests show that we can obtain relatively accurate locations and reliable uncertainties for earthquakes with 4 or 5 stations triggered. In particular, without considering the station latency, accurate earthquake locations can be achieved retrospectively in about 3.3 and 2.7 s for the 2003 San Simeon Mw6.6 earthquake and 2004 Parkfield Mw6.0 earthquake, respectively.