HR: 1340h
AN: S13A-1037    [Abstracts]
TI: Verification of the b-value model for moderate earthquakes in Kanto, Japan
AU: * IMOTO, M
EM: imoto@bosai.go.jp
AF: National Research Institute for Earth Science and Disaster Prevention, Tennodai 3-1, Tsukuba-shi, 305-0006 Japan
AU: Yamamoto, N
EM: n.yama@bosai.go.jp
AF: National Research Institute for Earth Science and Disaster Prevention, Tennodai 3-1, Tsukuba-shi, 305-0006 Japan
AB: In the Kanto region, decreases in b-values before earthquakes (M³5.5) have already been reported. A change of b-value is defined as the difference between the long-term average and short-term average of earthquake size. A hazard function of moderately large earthquakes incorporating this concept has been proposed (the b-value model). This model was based on the NIED catalogue of earthquakes that occurred between 1982 and 1999. The model is being subjected to verification testing based on earthquakes observed after January 2000 in order to evaluate the effectiveness of the model using independent data from both the NIED catalogue and the JMA catalogue. Although the accuracy of the b-value change may decrease, we will attempt to proceed with the test by expanding the study volume in order to increase the number of targets. The NIED catalogue records three earthquakes above the threshold magnitude of 5.5 as of Sept. 1. Decreases in b-value observed for these three earthquakes and an increase in the log-likelihood of the b-value model by 3.6 units, greater than that of the Poisson model, support the validity of the b-value model. The JMA catalogue records eight earthquakes above the same threshold and the same volume as those in the NIED catalogue. The difference in the number of targets between the two catalogues implies a systematic deviation of magnitude scale between them. To correct this bias, we derive an empirical relation converting the JMA magnitude into the NIED magnitude. By using the relation, the number of targets decreases to seven, and there is still a gap between them. We applied the b-model to the magnitude-converted JMA catalogue with the same model parameters. This test performed better than expected; the difference in the log-likelihood (information gain) was about 6. If we consider the performance for one event, information gain per event for the JMA case is about 0.8, which is a little smaller than that of NIED (1.2). The larger number of targets in the JMA case may include earthquakes of smaller magnitude than in the NIED catalogue, and this could result in a less effective gain in average than the NIED case. The results are consistent with each other and strongly support the b-value model.
DE: 7223 Seismic hazard assessment and prediction
DE: 7230 Seismicity and seismotectonics
SC: Seismology [S]
MN: 2004 AGU Fall Meeting