11:15 AM - 11:30 AM
[SSS11-14] Aleatory Variability of Seismic Ground Motion Based on a Non-ergodic Single-Site Ground Motion Model
Keywords:Aleatory Variability, Non-ergodic, Ground Motion Model, Single-Site
In Probabilistic Seismic Hazard Analysis (PSHA), it is crucial to account for epistemic uncertainty (due to limited data and knowledge) and aleatory variability (due to natural phenomena). The result of the analysis can vary significantly depending on how these uncertainties are handled. Aleatory variability remains even with additional data and knowledge, being specific to a location and not including the effects of epistemic uncertainty (e.g., from observation points, seismic wave propagation paths, or source locations). It is typically expressed as the standard deviation of a log-normal distribution, referred to as Single-Station Sigma (SSσ) when considering multiple earthquakes at a specific location, or Single-Path Sigma (SPσ) when propagation path effects are excluded.
With the recent accumulation of observational data, studies on SSσ and SPσ have advanced both domestically and internationally. Previous research includes investigations based on residuals between observation records and Ground Motion Models (GMMs) (e.g., Morikawa et al., 2013), studies not using GMMs with records from two earthquakes of the same magnitude and source (e.g., Hikita and Tomozawa, 2013), and studies based on residuals between observation records and Non-ergodic GMMs (NGMM) (e.g., Sung et al., 2023). These studies typically derive SSσ or SPσ by removing epistemic uncertainty from observational records at multiple sites rather than deriving single-site variability. Moreover, although some studies discuss the relationship between SSσ and factors like earthquake type or source distance, these relationships remain insufficiently explored due to data limitations and are not systematically organized. Following the 2011 Tohoku Earthquake, the assumed magnitude of earthquakes increased, and evaluations of ground motion near sources became essential, necessitating consideration of seismic source diversity. Clarifying the differences in variability between sites near and far from the source is crucial for improving large-scale and near-source ground motion evaluations.
In this study, seismic observation records from a single site in Japan were used to develop NGMMs for each location based on observations (PGA) from earthquakes in the same region, and the residuals between predicted and observed values were evaluated. To derive stable standard deviations at each location, records from Mw3.0 or higher earthquakes near the 2008 Iwate-Miyagi Inland Earthquake (reverse fault), 2013 Northern Ibaraki Earthquake (normal fault), and 2016 Kumamoto Earthquake (strike-slip fault) were used. SSσ was evaluated at K-NET and KiK-net observation sites where over 20 records were obtained. Si and Midorikawa (1999) was used as the underlying GMM, and records below 200 cm/s² were considered to avoid ground non-linearity effects.
SSσ was evaluated in three ways: (1) using Si and Midorikawa (1999) without limiting the source region for a single location, (2) using Si and Midorikawa (1999) with a limited source region, and (3) using NGMM with a limited source region. The NGMM was recalibrated using least squares to minimize residuals between observation records and predicted values, considering earthquake magnitude, shortest fault distance, depth, and ground amplification. The results showed a decrease in SSσ values in the order of (1), (2), and (3), demonstrating the effectiveness of NGMM in estimating SSσ. Furthermore, the relationship between fault distance, AVS30, and SSσ was examined for locations around the source regions of the three earthquakes. The results indicated a concentric decrease in variability with distance from the source region, with SSσ being larger (0.5 to 0.7) closer to the source and not exceeding 0.5 at distances greater than 30 km. SSσ values did not significantly vary with AVS30, and reverse and normal faults exhibited greater variability than strike-slip faults.
Future work will explore variability trends for other inland earthquakes, examining dependencies on various indicators and period-based trends.
Acknowledgments: The Strong-Motion Data Flatfile 2023 Edition from J-SHIS was used in this study.
With the recent accumulation of observational data, studies on SSσ and SPσ have advanced both domestically and internationally. Previous research includes investigations based on residuals between observation records and Ground Motion Models (GMMs) (e.g., Morikawa et al., 2013), studies not using GMMs with records from two earthquakes of the same magnitude and source (e.g., Hikita and Tomozawa, 2013), and studies based on residuals between observation records and Non-ergodic GMMs (NGMM) (e.g., Sung et al., 2023). These studies typically derive SSσ or SPσ by removing epistemic uncertainty from observational records at multiple sites rather than deriving single-site variability. Moreover, although some studies discuss the relationship between SSσ and factors like earthquake type or source distance, these relationships remain insufficiently explored due to data limitations and are not systematically organized. Following the 2011 Tohoku Earthquake, the assumed magnitude of earthquakes increased, and evaluations of ground motion near sources became essential, necessitating consideration of seismic source diversity. Clarifying the differences in variability between sites near and far from the source is crucial for improving large-scale and near-source ground motion evaluations.
In this study, seismic observation records from a single site in Japan were used to develop NGMMs for each location based on observations (PGA) from earthquakes in the same region, and the residuals between predicted and observed values were evaluated. To derive stable standard deviations at each location, records from Mw3.0 or higher earthquakes near the 2008 Iwate-Miyagi Inland Earthquake (reverse fault), 2013 Northern Ibaraki Earthquake (normal fault), and 2016 Kumamoto Earthquake (strike-slip fault) were used. SSσ was evaluated at K-NET and KiK-net observation sites where over 20 records were obtained. Si and Midorikawa (1999) was used as the underlying GMM, and records below 200 cm/s² were considered to avoid ground non-linearity effects.
SSσ was evaluated in three ways: (1) using Si and Midorikawa (1999) without limiting the source region for a single location, (2) using Si and Midorikawa (1999) with a limited source region, and (3) using NGMM with a limited source region. The NGMM was recalibrated using least squares to minimize residuals between observation records and predicted values, considering earthquake magnitude, shortest fault distance, depth, and ground amplification. The results showed a decrease in SSσ values in the order of (1), (2), and (3), demonstrating the effectiveness of NGMM in estimating SSσ. Furthermore, the relationship between fault distance, AVS30, and SSσ was examined for locations around the source regions of the three earthquakes. The results indicated a concentric decrease in variability with distance from the source region, with SSσ being larger (0.5 to 0.7) closer to the source and not exceeding 0.5 at distances greater than 30 km. SSσ values did not significantly vary with AVS30, and reverse and normal faults exhibited greater variability than strike-slip faults.
Future work will explore variability trends for other inland earthquakes, examining dependencies on various indicators and period-based trends.
Acknowledgments: The Strong-Motion Data Flatfile 2023 Edition from J-SHIS was used in this study.