14:30 〜 14:45
[STT38-04] 地震波と地震カタログの統合による本震直後の早期余震活動推定
★招待講演
キーワード：余震分布、検出確率、点過程、地震波
Large earthquakes frequently trigger a series of aftershock activities, which considerably reduce the signal-to-noise (SN) ratio in seismic waveforms and complicate aftershock detection. To address this issue, we propose a model that leverages detection probabilities and inferences from detected aftershock sequences. This model aims to mitigate the bias arising from the under-detection of aftershocks, a common obstacle in seismology.
However, modeling and estimating the detection function based on available data is inherently complex due to the dynamic nature of seismic events. Traditional methods have utilized a nonparametric detection function, estimated using earthquake catalog data. While these methods have been successful in estimating the global detection rate, they often fail to accurately capture local variations. This limitation underscores the need for more refined techniques capable of providing comprehensive insights into aftershock detection.
In this presentation, we introduce a cutting-edge method that employs seismic waveform data to inform detection rates directly. For instance, in Hi-net, seismic waveforms are sampled at a high frequency of 100Hz. Such detailed information is crucial for addressing the bias caused by the under-detection of aftershocks, allowing for a more accurate and complete understanding of early aftershock activity. However, integrating waveforms with catalog data presents a significant hurdle due to discrepancies in sampled time points. We propose a unique integrated approach that combines seismic waveforms and earthquake catalog data to overcome this hurdle. This method utilizes a Bayesian framework to efficiently estimate related parameters, providing a robust and comprehensive model for aftershock detection. Our approach not only facilitates a more accurate estimation of detection functions but also significantly improves parameters on the aftershock activity, such as b-value.
We have applied our model to the 2016 Kumamoto earthquake, demonstrating its ability to effectively capture both global and local detection rates. The results show estimated $b$-values of around 1.4, which is notably higher than the 0.8 computed using previous methods. This significant increase in b-values indicates the enhanced capability of our proposed method to detect small aftershocks.
Consequently, our method provides a more accurate and nuanced understanding of aftershock activities, contributing valuable insights into seismic research and earthquake preparedness.
However, modeling and estimating the detection function based on available data is inherently complex due to the dynamic nature of seismic events. Traditional methods have utilized a nonparametric detection function, estimated using earthquake catalog data. While these methods have been successful in estimating the global detection rate, they often fail to accurately capture local variations. This limitation underscores the need for more refined techniques capable of providing comprehensive insights into aftershock detection.
In this presentation, we introduce a cutting-edge method that employs seismic waveform data to inform detection rates directly. For instance, in Hi-net, seismic waveforms are sampled at a high frequency of 100Hz. Such detailed information is crucial for addressing the bias caused by the under-detection of aftershocks, allowing for a more accurate and complete understanding of early aftershock activity. However, integrating waveforms with catalog data presents a significant hurdle due to discrepancies in sampled time points. We propose a unique integrated approach that combines seismic waveforms and earthquake catalog data to overcome this hurdle. This method utilizes a Bayesian framework to efficiently estimate related parameters, providing a robust and comprehensive model for aftershock detection. Our approach not only facilitates a more accurate estimation of detection functions but also significantly improves parameters on the aftershock activity, such as b-value.
We have applied our model to the 2016 Kumamoto earthquake, demonstrating its ability to effectively capture both global and local detection rates. The results show estimated $b$-values of around 1.4, which is notably higher than the 0.8 computed using previous methods. This significant increase in b-values indicates the enhanced capability of our proposed method to detect small aftershocks.
Consequently, our method provides a more accurate and nuanced understanding of aftershock activities, contributing valuable insights into seismic research and earthquake preparedness.