講演情報
[U15-P04]気象庁震度データからの強地震動発生頻度予測・ユーザーによる解析を可能とする data dashboard の開発
*川西 琢也1 (1.金沢大学（現 EX Analytics）)
キーワード：
地震、極値理論、バックグラウンド・ポアソン強度、再現期間
The Japan Meteorological Agency (JMA) publishes nationwide records of ground shaking with seismic intensity >= 1 on its Shindo scale. Since April 1996, intensities have been reported as Instrumental Seismic Intensity (ISI) with 0.1 resolution, together with the rounded JMA Seismic Intensity Scale (SIS). For example, SIS 1 corresponds to ISI in the range [0.5, 1.4].
Kawanishi (2025) analyzed these data and found that, after removing temporal clustering such as aftershocks and swarms, the relationship between SIS and occurrence rates is strongly linear on semilog plots for SIS >= 2. This relationship also remains stable before and after major earthquakes, at least at stations that have recorded SIS 7. The study further showed that stations with higher estimated rates of SIS 6-plus shaking tend to experience large earthquakes more frequently. However, the estimated rates were often too high, and the large number of stations made detailed station-specific analysis difficult.
This presentation introduces two developments addressing these issues. First, we are creating an interactive website that allows users to select an observation station, set statistical parameters, and compute the risk of strong shaking at that site. The platform provides graphical diagnostics, including distributions of inter-shaking intervals, ISI values, and return-period curves. It is designed to support both academic research and practical applications such as disaster risk management. A prototype of the interface is attached to this abstract.
Second, we improve the previous methodology by using ISI instead of SIS to estimate the occurrence rate of large earthquakes, taking advantage of ISI's higher resolution.
We define a large event as SIS 6-plus (ISI >= 6.0). The occurrence rate is estimated as follows: (1) inter-shaking intervals of 10 days or more are fit to an exponential distribution to obtain the background Poisson intensity; (2) ISI values >= 0.5 follow a three-parameter gamma distribution, and the upper 10 percent of values are fit using extreme value theory to a Gumbel-type generalized Pareto distribution; (3) the background rate and tail model are combined to estimate the return period for SIS 6-plus shaking.
The resulting platform enables users to explore how modeling assumptions affect risk estimates and provides a practical tool for evaluating strong-motion hazards on a station-by-station basis.
Reference
Kawanishi, T. Estimating the risks of big earthquakes based solely on the empirical relationship between intensities and rates of occurrences after incorporating temporal clustering correction. Prog Earth Planet Sci 12, 25 (2025). https://doi.org/10.1186/s40645-025-00694-7
Kawanishi (2025) analyzed these data and found that, after removing temporal clustering such as aftershocks and swarms, the relationship between SIS and occurrence rates is strongly linear on semilog plots for SIS >= 2. This relationship also remains stable before and after major earthquakes, at least at stations that have recorded SIS 7. The study further showed that stations with higher estimated rates of SIS 6-plus shaking tend to experience large earthquakes more frequently. However, the estimated rates were often too high, and the large number of stations made detailed station-specific analysis difficult.
This presentation introduces two developments addressing these issues. First, we are creating an interactive website that allows users to select an observation station, set statistical parameters, and compute the risk of strong shaking at that site. The platform provides graphical diagnostics, including distributions of inter-shaking intervals, ISI values, and return-period curves. It is designed to support both academic research and practical applications such as disaster risk management. A prototype of the interface is attached to this abstract.
Second, we improve the previous methodology by using ISI instead of SIS to estimate the occurrence rate of large earthquakes, taking advantage of ISI's higher resolution.
We define a large event as SIS 6-plus (ISI >= 6.0). The occurrence rate is estimated as follows: (1) inter-shaking intervals of 10 days or more are fit to an exponential distribution to obtain the background Poisson intensity; (2) ISI values >= 0.5 follow a three-parameter gamma distribution, and the upper 10 percent of values are fit using extreme value theory to a Gumbel-type generalized Pareto distribution; (3) the background rate and tail model are combined to estimate the return period for SIS 6-plus shaking.
The resulting platform enables users to explore how modeling assumptions affect risk estimates and provides a practical tool for evaluating strong-motion hazards on a station-by-station basis.
Reference
Kawanishi, T. Estimating the risks of big earthquakes based solely on the empirical relationship between intensities and rates of occurrences after incorporating temporal clustering correction. Prog Earth Planet Sci 12, 25 (2025). https://doi.org/10.1186/s40645-025-00694-7