The discrimination filters to increase the reliability of EEW association on the location using geometric distribution of triggered stations with upgrading a travel time model.

Abstract:

In operation of Earthquake Early Warning System (EEWS), the alerting criteria are one of the most important parameters in optimizing acceptable warning system. During early stage of testing EEW systems from 2011 to 2013, we adapted ElarmS by UC Berkeley BSL to Korean seismic network and applied very simple criteria for event alerting with the combination of the numbers of station and magnitude. As a result of the testing we found out that the real-time test result of Earthquake Early Warning (EEW) system in Korea showed that all events located within seismic network with bigger than magnitude 3.0 were well detected. However, two events located at sea between land and island gave false results with magnitude over 4.0 related to the teleseismic waves and one event located in land gave false results with magnitude over 3.0 related to the teleseismic waves. These teleseismic-relevant false events were caused by logical co-relation during association procedure and the corresponding geometric distribution of associated stations is crescent-shaped. Seismic stations are not deployed uniformly, so the expected bias ratio varies with evaluated epicentral location. This ratio is calculated in advance and stored into database, called as TrigDB, for the discrimination of teleseismic-origin false alarm. We developed a method, so called ‘TrigDB back filling’, updating location with supplementary association of stations comparing triggered times between sandwiched stations which was not associated previously based on predefined criteria such as travel-time. Because EEW program assume that all events are local, teleseismic-relevant events can give more triggered stations by using back filling of the unassociated stations than the normal association. And we also developed a travel time curve (K-SEIS-1DTT2015) to reduce split event for EEWS. After applying the K-SEIS-1DTT2015 model, these teleseismic-relevant false events are reduced. As a result of these methods we could get more reliable location.