[S-CG53] 地震動・地殻変動・津波データの即時把握・即時解析・即時予測

[SCG53-P02] Constructing an empirical envelope function of seismic waveforms for the evaluation of EEW in Japan

*Hong Peng¹、Stephen Wu^1,2、Masumi Yamada³ (1.Research Organization of Information and Systems, The Institute of Statistical Mathematics, Tachikawa, 190-8562, Japan、2.The Graduate University for Advanced Studies, Department of Statistical Science, Tachikawa, 190-8562, Japan、3.Kyoto University, Research Division of Earthquake Disasters, Disaster Prevention Research Institute, Uji, 611-0011, Japan)

キーワード：Earthquake early warning, Real-time seismic waveforms, Envelope function

The earthquake early warning (EEW) system analyzes ground motion data from seismic networks and predicts the seismic intensity (SI) at locations where the strong ground motion has not arrived. Currently, the EEW system has been widely installed in many regions around the world (e.g., Japan and California). In Japan, there are two working EEW systems: integrated particle filter (IPF) and propagation of local undamped motion (PLUM). IPF predicts the SI through seismic source inversion, which is different from PLUM that directly predicts the SI by using waveform propagation theory on real-time seismic waveform data. As a result, IPF is generally faster but less accurate than PLUM. Regarding the hazard assessment, the Japan Meteorological Agency (JMA) will select the largest SI between the two systems above and broadcasts a single warning to the public. In this study, we propose a new evaluation method that can compare the performance of the two systems in real-time. First, we collect the seismic event data from JMA, Kyoshin Network (K-NET), and Kiban Kyoshin Network (KiK-NET) in the past 20 years, building hitherto the largest benchmark database. Then, we convert the seimisc waveforms to SI curves and construct an empirical envelope function. Finally, the evaluation method comes from comparing the similarity between the predicted envelopes from EEW and observed envelopes in real-time. The effectiveness of our method will be tested on real seismic events in Japan.