10:45 AM - 11:00 AM
[SSS06-07] Toward the construction of an integrated assessment method of crustal activity -characterization of “normal” seismicity in Japan-
Keywords:seismicity, tidal correlation, ETAS, starain rate
Meteorological Research Institute is progressing research to develop the method to assess the crustal activity such as seismicity and crustal deformation by integrating the various parameters that characterize the activities. The ultimate goal is to assess the potential of the integrated method to express the current anomaly of crustal activity. For that purpose, it is necessary to quantify the “normal” crustal activity from a more integrated perspective not only from the conventional viewpoint based on the results of individual parameters. In the presentation, we will report the result of the analysis of the joint distribution of the parameters characterizing the size distribution, tidal correlation, and activation/quiescence of seismicty estimated from hypocenter data for the past 20 years in Japan conducted as an attempt to develop the integrated assessment method.
In the analysis, we used the JMA unified seismic catalog from 2000 to August 2021. We extracted two types of dataset consist of shallow-inland earthquakes (M ≧ 2.0, depth ≦30 km) and earthquakes including subduction-zone earthquakes (M≧3.5, depth≦100 km) from the catalog. For each dataset, the parameter values are estimated for the same N consecutive earthquakes in dl degree square spatial grids. The results are smoothed by shifting the analysis window by half, that is, N/2 in the temporal direction and dl/2 in the spatial direction. The results give us a joint distribution in addition to the individual distributions of each parameter. When comparing the distribution with parameter values observed in specific time or space, it is possible to apply a conventional statistical test such as the goodness-of-fit test by extracting the results of grids that do not overlap each other.
As a parameter characterizing the size distribution of seismicity, the GR law’s b value was estimated. The earthquakes that clearly deviate from the GR law, such as aftershocks just after a large earthquake, were excluded by the R value (Wiemer and Wyss 2000, 2002). As a parameter characterizing the tidal correlation of seismicity, based on Schuster's test (Schuster 1897), we used the D value (D2=(Σi=1N cosθi)2+ (Σi=1N sinθi)2 , where θi is the tidal phase angle of i'th earthquake) to estimate the correlation between seismicity and phase angle of tidal response of volumetric strain. To eliminate the apparent correlation caused by swarming earthquakes, instead of using declustered data, which makes difficult to estimate joint distribution, we excluded the case where N/4 or more events occurred within half of the tidal cycle. For characterizing activation/quiescence of seismicity, the expected number of earthquakes in the analysis window by the ETAS model (E value, hereafter), whose parameters were estimated in each spatial grid, is used. Some of the ETAS parameters used here can also be used as indicators of seismicity’s regionality.
The probability density distributions of the three parameters (b, D, E) are close to, but different from those expected from theory and model. Though the difference should include the spatiotemporal fluctuations that should be investigated in detail, it makes sense to quantify them as an average characteristic of "normal" seismic activity by excluding the known "abnormal" cases such as the very early stage of aftershocks where the frequency distribution clearly deviates from the GR law or where earthquakes occur too frequently to estimate tidal correlation. With the joint distributions of these parameters, for example, when anomalies in the size distribution and tidal correlation of seismicity occur simultaneously, it will be possible to quantify how abnormal the observed phenomena are. It is also important to know whether a distribution of the parameter values different from "normal" may be obtained before the occurrence of a significant earthquake or in response to crustal deformations. In the presentation, we would like to introduce some examples from this perspective.
In the analysis, we used the JMA unified seismic catalog from 2000 to August 2021. We extracted two types of dataset consist of shallow-inland earthquakes (M ≧ 2.0, depth ≦30 km) and earthquakes including subduction-zone earthquakes (M≧3.5, depth≦100 km) from the catalog. For each dataset, the parameter values are estimated for the same N consecutive earthquakes in dl degree square spatial grids. The results are smoothed by shifting the analysis window by half, that is, N/2 in the temporal direction and dl/2 in the spatial direction. The results give us a joint distribution in addition to the individual distributions of each parameter. When comparing the distribution with parameter values observed in specific time or space, it is possible to apply a conventional statistical test such as the goodness-of-fit test by extracting the results of grids that do not overlap each other.
As a parameter characterizing the size distribution of seismicity, the GR law’s b value was estimated. The earthquakes that clearly deviate from the GR law, such as aftershocks just after a large earthquake, were excluded by the R value (Wiemer and Wyss 2000, 2002). As a parameter characterizing the tidal correlation of seismicity, based on Schuster's test (Schuster 1897), we used the D value (D2=(Σi=1N cosθi)2+ (Σi=1N sinθi)2 , where θi is the tidal phase angle of i'th earthquake) to estimate the correlation between seismicity and phase angle of tidal response of volumetric strain. To eliminate the apparent correlation caused by swarming earthquakes, instead of using declustered data, which makes difficult to estimate joint distribution, we excluded the case where N/4 or more events occurred within half of the tidal cycle. For characterizing activation/quiescence of seismicity, the expected number of earthquakes in the analysis window by the ETAS model (E value, hereafter), whose parameters were estimated in each spatial grid, is used. Some of the ETAS parameters used here can also be used as indicators of seismicity’s regionality.
The probability density distributions of the three parameters (b, D, E) are close to, but different from those expected from theory and model. Though the difference should include the spatiotemporal fluctuations that should be investigated in detail, it makes sense to quantify them as an average characteristic of "normal" seismic activity by excluding the known "abnormal" cases such as the very early stage of aftershocks where the frequency distribution clearly deviates from the GR law or where earthquakes occur too frequently to estimate tidal correlation. With the joint distributions of these parameters, for example, when anomalies in the size distribution and tidal correlation of seismicity occur simultaneously, it will be possible to quantify how abnormal the observed phenomena are. It is also important to know whether a distribution of the parameter values different from "normal" may be obtained before the occurrence of a significant earthquake or in response to crustal deformations. In the presentation, we would like to introduce some examples from this perspective.