5:15 PM - 6:45 PM
[SSS10-P04] Extraction of Near-fault Pulse-like Ground Motion based on the Empirical Mode Decomposition : Analysis of the 2023 Turkey-Syria Earthquake (Mw7.8)
Keywords:pulse-like ground motion, 2023 Turkey-Syria Earthquake, Empirical Mode Decomposition
Introduction
Strong pulse-like ground motion is observed near faults and is known to cause significant damage to structures. To objectively extract pulse-like ground motion from near-fault velocity seismograms, a wavelet-transform-based method has been widely used (e.g., Baker, 2007). Chen et al. (2019) proposed a method for extracting pulse-like ground motion using the Empirical Mode Decomposition (EMD) that decomposes a waveform into several oscillation modes called Intrinsic Mode Functions (IMF). On February 6, 2023, a Mw7.8 left-lateral strike-slip earthquake occurred in Turkey's central and southern parts. Wu et al. (2023) already applied Baker's method (2007) to the observed waveforms of this earthquake and extracted pulses. However, the method can extract only one pulse at a time. Therefore, in this study, we applied the EMD-based method (Chen et al., 2019; Chen et al., 2020), which can extract multiple pulses, to the observed waveforms of the Mw7.8 earthquake.
Method
We used acceleration waveforms (north-south and east-west components) at 13 stations near the East Anatolian Fault. After converting acceleration waveforms to velocity waveforms using numerical integration, baseline correction was performed using a 30-second median filter (Watanabe et al., 2021). Subsequently, the waveforms were rotated in the horizontal plane from 0 to 360 degrees every 10 degrees, and the EMD-based method was applied to extract pulses (rough pulses) for each direction. The pulses were extracted by selecting IMFs that met the criteria of PGV/PGA ratio and energy contribution. The extracted pulses were evaluated using the Pulse Indicator (Baker, 2007) to represent the pulse intensity of the waveforms, and the orientation of the pulses at each station was determined. Then, half-cycles with more than 5% of the total energy were extracted as inherent pulses, and the period was calculated from the time difference of zero crossings.
Results
Directivity pulses were extracted from waveforms perpendicular to the fault strike at many stations, resulting in a significant increase in the Pulse Indicator value. On the other hand, at stations especially near the fault, pulses were also extracted from waveforms parallel to the fault strike due to fling steps. Only one pulse was extracted at most of the stations. On the other hand, two pulses were extracted at station 4615. Actually, another wave packet was found at this station, but was not identified to be a pulse according to our current threshold. Then, we compared the period and PGV of extracted pulses in the fault-normal direction (determined as the direction perpendicular to the direction of the maximum displacement) with existing scaling laws. A scaling law between the earthquake magnitude and the pulse period (Chen et al., 2020) generally matched. However, the relation between fault distance and PGV in this study was slightly smaller compared to previous scaling laws (Chen et al., 2020; Shi and Midorikawa, 1999). There was no clear relationship between fault distance and pulse period.
Discussion and Conclusion
For each of the two pulses extracted at station 4615, the displacement was determined. The first pulse has the maximum displacement in the north-south direction, and the second pulse has the maximum displacement in the northeast-southwest direction. Station 4615 is located near two fault segments. The strike directions of the first and second fault segments are generally oriented in the north-south and northeast-southwest directions, respectively. Therefore, it is considered that the two pulses were generated by the different fault segments. We have confirmed that the EMD-based method can extract multiple pulses.
[Acknowledgement]
We used the strong motion records of the Disaster and Emergency Management Authority (AFAD).
Strong pulse-like ground motion is observed near faults and is known to cause significant damage to structures. To objectively extract pulse-like ground motion from near-fault velocity seismograms, a wavelet-transform-based method has been widely used (e.g., Baker, 2007). Chen et al. (2019) proposed a method for extracting pulse-like ground motion using the Empirical Mode Decomposition (EMD) that decomposes a waveform into several oscillation modes called Intrinsic Mode Functions (IMF). On February 6, 2023, a Mw7.8 left-lateral strike-slip earthquake occurred in Turkey's central and southern parts. Wu et al. (2023) already applied Baker's method (2007) to the observed waveforms of this earthquake and extracted pulses. However, the method can extract only one pulse at a time. Therefore, in this study, we applied the EMD-based method (Chen et al., 2019; Chen et al., 2020), which can extract multiple pulses, to the observed waveforms of the Mw7.8 earthquake.
Method
We used acceleration waveforms (north-south and east-west components) at 13 stations near the East Anatolian Fault. After converting acceleration waveforms to velocity waveforms using numerical integration, baseline correction was performed using a 30-second median filter (Watanabe et al., 2021). Subsequently, the waveforms were rotated in the horizontal plane from 0 to 360 degrees every 10 degrees, and the EMD-based method was applied to extract pulses (rough pulses) for each direction. The pulses were extracted by selecting IMFs that met the criteria of PGV/PGA ratio and energy contribution. The extracted pulses were evaluated using the Pulse Indicator (Baker, 2007) to represent the pulse intensity of the waveforms, and the orientation of the pulses at each station was determined. Then, half-cycles with more than 5% of the total energy were extracted as inherent pulses, and the period was calculated from the time difference of zero crossings.
Results
Directivity pulses were extracted from waveforms perpendicular to the fault strike at many stations, resulting in a significant increase in the Pulse Indicator value. On the other hand, at stations especially near the fault, pulses were also extracted from waveforms parallel to the fault strike due to fling steps. Only one pulse was extracted at most of the stations. On the other hand, two pulses were extracted at station 4615. Actually, another wave packet was found at this station, but was not identified to be a pulse according to our current threshold. Then, we compared the period and PGV of extracted pulses in the fault-normal direction (determined as the direction perpendicular to the direction of the maximum displacement) with existing scaling laws. A scaling law between the earthquake magnitude and the pulse period (Chen et al., 2020) generally matched. However, the relation between fault distance and PGV in this study was slightly smaller compared to previous scaling laws (Chen et al., 2020; Shi and Midorikawa, 1999). There was no clear relationship between fault distance and pulse period.
Discussion and Conclusion
For each of the two pulses extracted at station 4615, the displacement was determined. The first pulse has the maximum displacement in the north-south direction, and the second pulse has the maximum displacement in the northeast-southwest direction. Station 4615 is located near two fault segments. The strike directions of the first and second fault segments are generally oriented in the north-south and northeast-southwest directions, respectively. Therefore, it is considered that the two pulses were generated by the different fault segments. We have confirmed that the EMD-based method can extract multiple pulses.
[Acknowledgement]
We used the strong motion records of the Disaster and Emergency Management Authority (AFAD).