JpGU-AGU Joint Meeting 2017

Presentation information

[EE] Oral

S (Solid Earth Sciences) » S-SS Seismology

[S-SS08] [EE] Earthquake Modeling and Simulation

Mon. May 22, 2017 9:00 AM - 10:30 AM A07 (Tokyo Bay Makuhari Hall)

convener:Eiichi Fukuyama(National Research Institute for Earth Science and Disaster Prevention), John B Rundle(University of California Davis), Yukitoshi Fukahata(Disaster Prevention Research Institute, Kyoto University), Chairperson:John Rundle(UC Davis), Chairperson:Eiichi Fukuyama(NIED)

9:00 AM - 9:15 AM

[SSS08-01] Test of the predictability of PI method on the Tohoku Mw9.0 earthquake

*Yongxian Zhang1, Cheng Song2, Caiyun Xia3, Shengfeng Zhang4 (1.China Earthquake Networks Center, 2.Institute of Earthquake Science, China Earthquake Administration, 3.Liaoning Earthquake Administration, 4.Institute of Geophysics, China Earthquake Administration)

Keywords:PI method, Tohoku Mw9.0 earthquake, predictability, ROC test, R score test

In this research, the local area (32.0°~46.0°N, 136.0°~148.0°E) including most of Japan was chosen to be the study region for verifying the predictability of the pattern informatics (PI) method under different models with different parameters using the receiver-operating characteristic (ROC) curve test and R score test. Pattern Informatics (PI) method was applied to the retrospective study on the forecasting of large earthquakes especially the Tohoku Mw9.0 earthquake in this region. Different forecasting maps with different calculating parameters were obtained. The main calculating parameters were respectively the grid size of 0.5°×0.5° or 1.0°×1.0° and forecasting window lengths from 5 to 10 years. The results showed that in most of the models, the hotspots were in its Moore neighborhood grids or its epicentral grid in the forecasting windows containing the Mw9.0 Tohoku earthquake, which suggests that the PI method could forecast the Tohoku Mw9.0 earthquake. The results also showed that under the ROC test and R score test the models with larger grid size (1.0°×1.0°) and longer forecasting window length (7~10 years), the forecasting effect were better.