18:15 〜 19:30
[STT57-P06] 地震波自動処理システムの性能評価（２）
キーワード：地震波初動自動検測, 地震波自動処理システム, 性能評価
The number of seismic stations has tremendously increased by many temporary seismic networks recently deployed in various areas, in addition to dense routine seismic networks such as the nationwide Kiban seismic network. Effective automatic earthquake detection and location system is anticipated, because the ability of data processing is limited. Manually picking P- and S-wave arrival times etc. from a huge amount of seismic waveform data observed by such many seismic stations is considerably time consuming work.
Horiuchi et al. (2012, 2013) have developed such an automatic seismic waveform processing system. This system was set up at Tohoku University on December 2012, and automatic detection and location processing of the nationwide seismic network data has been operating since then. The system can detect and locate many earthquakes which are difficult to be located by the routine processing based on manual pickings. However, sometimes earthquakes cannot be correctly discriminated by the system: for example, when more than two earthquakes occur almost simultaneously. In order to consider the application of automatic earthquake detection and location system to the actual seismic network, we need to know its performance.
Nakayama et al. (2013) tried to evaluate performance of this earthquake detection and location system for the application to the nationwide seismic network. Results showed that the automatic system could detect and locate earthquakes about 1.5 times more than those in the JMA unified catalogue. The automatic system extended the lower limit of the detection capability to much smaller magnitude range than that by the JMA unified catalogue. The evaluation also showed that S-wave arrival times picked by the automatic system were systematically delayed by ~0.05-0.1 sec compared with those by the manual pickings of the unified catalogue. Based on this performance evaluation, Horiuchi et al. (2014 this meeting) have tried to improve the system by developing a new algorithm to better pick S-wave arrivals.
We have evaluated performance of this presently improved automatic processing system by using the waveform data for the same period as those in the previous evaluation. Results show that the systematic delay of S-wave arrivals by the automatic pickings is considerably improved and the difference in S-wave arrivals between the new automatic system and the unified catalogue has become nearly the same as that between the manual pickings by Tohoku University and those in the unified catalogue. This indicates that the S-wave arrival times, as well as P-wave arrival times, picked by the automatic system almost stand comparison with those by the manual picking. Moreover, the evaluation shows that the new system also improved the rate of correct discrimination of earthquakes: the percentage of events that were missed to be correctly located decreased from 19% to 14% (most of these events are those located in and around the Izu-Bonin Islands and the Ryukyu Islands), and the percentage of events that were incorrectly defined as earthquakes decreased from 3.1% to 2.5%. This is because of the improvement of algorithm to correctly discriminate more than two earthquakes that occurred nearly simultaneously.
Horiuchi et al. (2012, 2013) have developed such an automatic seismic waveform processing system. This system was set up at Tohoku University on December 2012, and automatic detection and location processing of the nationwide seismic network data has been operating since then. The system can detect and locate many earthquakes which are difficult to be located by the routine processing based on manual pickings. However, sometimes earthquakes cannot be correctly discriminated by the system: for example, when more than two earthquakes occur almost simultaneously. In order to consider the application of automatic earthquake detection and location system to the actual seismic network, we need to know its performance.
Nakayama et al. (2013) tried to evaluate performance of this earthquake detection and location system for the application to the nationwide seismic network. Results showed that the automatic system could detect and locate earthquakes about 1.5 times more than those in the JMA unified catalogue. The automatic system extended the lower limit of the detection capability to much smaller magnitude range than that by the JMA unified catalogue. The evaluation also showed that S-wave arrival times picked by the automatic system were systematically delayed by ~0.05-0.1 sec compared with those by the manual pickings of the unified catalogue. Based on this performance evaluation, Horiuchi et al. (2014 this meeting) have tried to improve the system by developing a new algorithm to better pick S-wave arrivals.
We have evaluated performance of this presently improved automatic processing system by using the waveform data for the same period as those in the previous evaluation. Results show that the systematic delay of S-wave arrivals by the automatic pickings is considerably improved and the difference in S-wave arrivals between the new automatic system and the unified catalogue has become nearly the same as that between the manual pickings by Tohoku University and those in the unified catalogue. This indicates that the S-wave arrival times, as well as P-wave arrival times, picked by the automatic system almost stand comparison with those by the manual picking. Moreover, the evaluation shows that the new system also improved the rate of correct discrimination of earthquakes: the percentage of events that were missed to be correctly located decreased from 19% to 14% (most of these events are those located in and around the Izu-Bonin Islands and the Ryukyu Islands), and the percentage of events that were incorrectly defined as earthquakes decreased from 3.1% to 2.5%. This is because of the improvement of algorithm to correctly discriminate more than two earthquakes that occurred nearly simultaneously.