Validation and uncertainties of strong ground motion prediction methods for seismic hazard assessment: a case study of the 2024 Noto Peninsula earthquake (Mw7.5), Japan

Abstract The major active fault earthquakes in Japan in recent years include the 2016 Kumamoto Earthquake and the 2024 Noto Peninsula earthquake. These earthquakes occurred along predicted fault models, however, uncertainties and variations in fault geometry and rupture propagation led to actual earthquake magnitudes exceeding prior expectations. The objective of this study is to verify if the Noto Peninsula earthquake could have been predicted, considering significant uncertainties in general strong motion prediction models, and to obtain feedback leading to improvements. Therefore, this study set a fault model using data from before the earthquake and performed a retrospective prediction analysis. This was compared with observed data to confirm predictability. Considering uncertainties in fault parameters and rupture propagation patterns, strong motion was generally predictable, except for the west coast and south coast of the Noto Peninsula. Subsequently, the uncertainties encompassed by this study was examined through the Tornado Plots for each calculation point, with a particular emphasis on the factors exerting the most substantial influence. The “Rupture Propagation Patterns” were the primary factor, with the “Concept of Fault Width” being the second most influential factor in many locations. However, at some locations near the fault top, the “Rake Angle” was shown to have a significant impact. This suggests that the effects of fault parameter uncertainties may exhibit regional characteristics. In the standard seismic hazard assessment methodology in Japan, the asperity model is utilized among characteristic source models. The asperity model effectively reproduces the pulsed waves near the source fault that can be represented by a simple model with a few asperities on a single fault plane. However, it is difficult to reproduce waveforms involving multiple fault ruptures and complex source processes, such as those seen in the Noto Peninsula earthquake. This study considers the limitations of applying characteristic source models. It proposes that, rather than aiming for exact waveform matching, evaluations should be permitted under loose criteria. Furthermore, by allowing for a certain degree of variability not only in fault parameter uncertainties but also in the calculation results themselves, such models can be considered suitable for use in assessments.