Yi-Hsuan Wu

Precursory Migration of Anomalous Seismic Activity Revealed by the Pattern Informatics Method: A Case Study of the 2011 Tohoku Earthquake, Japan

Anomalous seismicity patterns in epicentral and surrounding areas have been shown to occur prior to large earthquakes, although the processes determining the spatial distribution and migration patterns of such seismicity are still poorly understood. We applied the improved pattern informatics (PI) method to earthquake data maintained by the Japan Meteorological Agency (JMA) for a broad region including northeastern inland Japan and the source area of the 2011 Tohoku earthquake, in order to reveal the precursory processes of the earthquake and its related statistical features. In particular, we focused on the spatial distribution and migration patterns of PI hotspots, which highlight areas of anomalous seismic activity. Our results show that such hotspots had been approaching the epicenter of the 2011 Tohoku earthquake from 2000 until the earthquake occurred. The possibility that this result was obtained by chance was rejected at the 95% confidence level based on Molchan’s error diagram. Our result supports the hypothesis that the preparatory processes of the 2011 Tohoku earthquake included anomalous seismic activity and its migration toward the epicenter of the earthquake.

Critical parameter estimates for earthquake forecast using PI migration

Parameters usually play a key role in statistical forecasting models and should be carefully determined because the physics of the system links to the statistical model through the parameters. To investigate the relations between parameters and pattern informatics (PI) migration, we developed a series of retrospective analyses. The results show that two parameters (i.e., cut magnitude and change interval) are essential factors of calculating PI migration. The cut magnitude is a lower cutoff magnitude applied to the catalog at the start of the analysis, and the result of the analysis shows that the PI migration hot spots are mostly distributed around the earthquakes with magnitude larger than target magnitude when the cut magnitude is 3.2–3.4 in most study regions. The change interval is a time span prior to the large event that we assumed to be the duration of the preparation process. In the retrospective analysis, the ability of PI migration hot spot to hit the target earthquake varies with change interval and the change interval that make PI migration hot spot to hit most target earthquakes varies with the study region. By using a retrospective analysis, we determined the optimal parameters for each study region, generating PI migration maps to show potential locations.