5:15 PM - 7:15 PM
[SSS11-P19] Effects of uncertainties in earthquake source characterization and ground motion modeling on probabilistic seismic hazard analysis (PSHA)
Keywords:Probabilistic Seismic Hazard Analysis, Uncertainty, Earthquake Occurrence Modeling, Ground Motion Prediction Equation, Attenuation Relationships of Ground Motion
The following two studies were conducted to capture the impact of uncertainties in earthquake occurrence modeling and ground motion modeling on PSHA.
1. Effects of uncertainties in earthquake occurrence modeling of active faults on PSHA
In this study, we conducted PSHA to analyze the effects of uncertainties in estimating magnitudes and occurrence probabilities of earthquakes caused by an active fault in the northern part of the Itoigawa-Shizuoka Tectonic Line. We used magnitudes and calculated earthquake occurrence probabilities based on the characteristic model adopted by the Headquarters for Earthquake Research Promotion (HERP1). We also estimated the earthquake magnitude and occurrence probability using a different characteristic earthquake model, YC85 model2. The HERP model deals with the magnitude and occurrence probability of a large characteristic earthquake. In contrast, the YC85 model takes into account the magnitude of small and medium earthquakes in addition to the uncertainty in the magnitude of large earthquakes occurring on the fault. We conducted PSHAs with the HERP and YC85 models at several fault distances in terms of peak ground acceleration (PGA) and response spectrum and analyzed the differences in the results.
As the results, we found that the PGA from the latter model is larger at high level of annual exceedance probability, and the effects of small and medium-sized earthquakes are evident, but at low probabilities, the two models are almost equivalent. In addition, when the regional seismic source is also taken into account for both models, we found that there is no significant difference in the hazard curves.
2. The effects of uncertainties in ground motion prediction on PSHA
First, we estimated the site-specific amplification property (αS) and event-specific source property (βe) from the residuals between observed and predicted ground motions and corrected the median value of an existing Ground Motion Prediction Equation (GMPE)3 using αS to match the observed ground motions at a specific single station. The standard deviation at the single station was estimated from the residuals between the observed ground motions and those predicted by the corrected GMPE. Note that the observed ground motions are free-field motions at the surface of a hard ground, analytically estimated from the records of the KiK-net borehole stations (Vs ~400 to 3300 m/s). When the αS is used to correct two existing GMPEs, we found that the predicted median values are generally closer to the observed values and the standard deviations are reduced for both GMPEs. Depending on the GMPE used for correction, however, there are some differences among the predicted ground motions after correction, and the correction may result in a larger deviation and/or median values than the original GMPE prediction at some stations.
Next, PSHA using the GMPE before and after the above correction was performed for three sites with different Vs values at the borehole seismic stations, and uniform hazard spectra (UHS) were compared with annual exceedance probabilities of 10-6 to 10-4. The results showed that the UHS using the corrected equation tended to be significantly smaller than those using the uncorrected equation because of smaller standard deviations, but the effect of the median correction was also observed. In the PSHA, therefore, it is important to use multiple GMPEs and correct not only the standard deviation but also the median to appropriately take uncertainty into account.
References
1. HERP, https://www.jishin.go.jp/evaluation/long_term_evaluation/
2. Youngs & Coppersmith, Bull. Seism. Soc. Amer. ,75, 939-964.
3. Morikawa & Fujiwara, J. Disas. Res., 8, 878-888.
1. Effects of uncertainties in earthquake occurrence modeling of active faults on PSHA
In this study, we conducted PSHA to analyze the effects of uncertainties in estimating magnitudes and occurrence probabilities of earthquakes caused by an active fault in the northern part of the Itoigawa-Shizuoka Tectonic Line. We used magnitudes and calculated earthquake occurrence probabilities based on the characteristic model adopted by the Headquarters for Earthquake Research Promotion (HERP1). We also estimated the earthquake magnitude and occurrence probability using a different characteristic earthquake model, YC85 model2. The HERP model deals with the magnitude and occurrence probability of a large characteristic earthquake. In contrast, the YC85 model takes into account the magnitude of small and medium earthquakes in addition to the uncertainty in the magnitude of large earthquakes occurring on the fault. We conducted PSHAs with the HERP and YC85 models at several fault distances in terms of peak ground acceleration (PGA) and response spectrum and analyzed the differences in the results.
As the results, we found that the PGA from the latter model is larger at high level of annual exceedance probability, and the effects of small and medium-sized earthquakes are evident, but at low probabilities, the two models are almost equivalent. In addition, when the regional seismic source is also taken into account for both models, we found that there is no significant difference in the hazard curves.
2. The effects of uncertainties in ground motion prediction on PSHA
First, we estimated the site-specific amplification property (αS) and event-specific source property (βe) from the residuals between observed and predicted ground motions and corrected the median value of an existing Ground Motion Prediction Equation (GMPE)3 using αS to match the observed ground motions at a specific single station. The standard deviation at the single station was estimated from the residuals between the observed ground motions and those predicted by the corrected GMPE. Note that the observed ground motions are free-field motions at the surface of a hard ground, analytically estimated from the records of the KiK-net borehole stations (Vs ~400 to 3300 m/s). When the αS is used to correct two existing GMPEs, we found that the predicted median values are generally closer to the observed values and the standard deviations are reduced for both GMPEs. Depending on the GMPE used for correction, however, there are some differences among the predicted ground motions after correction, and the correction may result in a larger deviation and/or median values than the original GMPE prediction at some stations.
Next, PSHA using the GMPE before and after the above correction was performed for three sites with different Vs values at the borehole seismic stations, and uniform hazard spectra (UHS) were compared with annual exceedance probabilities of 10-6 to 10-4. The results showed that the UHS using the corrected equation tended to be significantly smaller than those using the uncorrected equation because of smaller standard deviations, but the effect of the median correction was also observed. In the PSHA, therefore, it is important to use multiple GMPEs and correct not only the standard deviation but also the median to appropriately take uncertainty into account.
References
1. HERP, https://www.jishin.go.jp/evaluation/long_term_evaluation/
2. Youngs & Coppersmith, Bull. Seism. Soc. Amer. ,75, 939-964.
3. Morikawa & Fujiwara, J. Disas. Res., 8, 878-888.