[S-TT41] Seismic monitoring and processing system

[STT41-P06] Development of the rapid moment tensor estimation system for seismic monitoring in Thailand

*Mongkolchai Sukmee¹, Hiroyuki Kumagai¹ (1.Graduate School of Environmental Studies, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464 8601, Japan)

The Thai Meteorological Department (TMD) started to operate the seismic observation network to monitor seismic activity in 2007 after the 2004 Sumatra earthquake. Nowadays, the Thai network has capabilities to estimate the magnitudes, origin times, and locations of earthquakes in Thailand and nearby countries but still has no real-time moment tensor estimation system. The moment tensor is fundamental to estimating the earthquake source mechanism and predicting tsunami heights. We developed the rapid moment tensor estimation by adopting the SWIFT and SWIFT TSUNAMI systems to the Thai network. SWIFT (Source parameter determination based on Waveform Inversion of Fourier Transformed seismograms) was developed by Nakano et al. (2008) to estimate the moment function and centroid moment tensor (CMT) for large earthquakes in real-time. SWIFT TSUNAMI developed by Inazu et al. (2016) is a tsunami forecast system based on the SWIFT CMT solution. To evaluate the reliability of the SWIFT system with seismic data in Thailand, we manually estimate the CMT solutions of moderate to large earthquakes in the TMD catalog between 2011 and 2021 by using data from the TMD archive and compared them with CMT solutions from the Global CMT Project (GCMT) catalog. We obtained 55 SWIFT CMT solutions (4.4 ≦ M_w ≦ 6.8). We used the Kagan angle (Kagan 2007) to compare the different angles or rotation angles between SWIFT CMT and GCMT solutions. The seismic moments, moment magnitudes, focal mechanisms, and source locations of SWIFT CMT were consistent with those of GCMT. The average moment magnitude difference was -0.002 with a standard deviation of 0.14. The Kagan angle was less than 50° for more than 80% of all events. Most of the manual SWIFT solutions are located in Myanmar, the Andaman Islands, and Indonesia. These areas can release energy into strong earthquakes that can affect Thailand. For that reason, we extended the area of study to Myanmar, the Andaman Islands, and western Indonesia by using seismic data from the nearby international and national networks. SWIFT automatically selects waveforms by using signal-to-noise amplitude ratio (S/N), source amplitude ratio (R) (Sakai et al. 2016), and kurtosis values (γ). We determined those values from manual CMT solutions between 2011 and 2022. The best values for automatic waveform selection for Thai and nearby networks are S/N more than 3, R less than 11, and γ greater than 2.5. After we evaluated the reliability and adjusted the system, we applied SWIFT to the real-time data from the Thai and nearby networks. SWIFT requires the initial hypocenter location of the earthquake to start the inversion process. For real-time earthquakes, we used the SeisComP system to estimate the initial hypocenter location, magnitude, and event time. We obtained automatic CMT solutions of 45 earthquakes between August and December 2022 in real time. We compared the automatic with the manual SWIFT CMT solutions to evaluate the reliability. The automatic SWIFT CMT solutions were highly consistent with those of manual SWIFT CMT solutions. The average moment magnitude difference was -0.08 with a standard deviation of 0.22. The Kagan angle was less than 40° for more than 70% of all events. SWIFT TSUNAMI requires the CMT solution of SWIFT to start the process which takes approximately 5 to 10 minutes of waiting. A rapid tsunami forecast of a large earthquake is necessary for tsunami warnings and evacuation. To fill this gap, we developed a preliminary tsunami forecast system by adapting SWIFT TSUNAMI with the database search CMT solution. We used the CMT solutions from the GCMT catalog between 1976 and 2021 as a database and selected the most possible CMT solution for each event from a database based on the estimated hypocenter location and magnitude from SeisComP system. We compared the database search CMT solutions with the manual SWIFT CMT solutions to evaluate their reliability. The focal mechanisms of the database search solutions were consistent with those of the SWIFT CMT solutions for large earthquakes (M_w ≧ 6.0). The modified SWIFT TSUNAMI can provide a preliminary tsunami forecast based on the database search CMT solution within three minutes which will be useful for warnings purpose. SWIFT and SWIFT TSUNAMI showed good performance in automatically estimating the moment tensors and forecasting tsunamis in Thailand and nearby regions in the prompt time. This pilot system for rapid moment tensor estimation will be used to improve the monitoring of seismic activities in and around Thailand.