3:00 PM - 3:15 PM
[SSS12-06] Earthquake Prediction and Diagnostic Analysis Using the Hierarchical Space-Time ETAS (HIST-ETAS) Model
Keywords:Delaunay Triangulation, Repeating earthquakes, Akaike Bayesian Information Criterion, Maximum a Posteriori solution, Penalized log likelihood, Information gain (log-likelifood ratio)
Overview: The HIST-ETAS model is a detailed location-dependent model designed to predict when and where earthquakes might occur across Japan, including specific regions. This model forecasts future earthquake occurrence rates based on past earthquake data. It utilizes the background seismicity rate to predict the location of major earthquakes.
Background Seismicity: The HIST-ETAS model separates and extracts the typical seismic activity, using its intensity (background seismicity rate) for predictions. The intensity derived from the Japan Meteorological Agency catalog for nationwide (M>4.95) and inland (M>3.95) regions shows good correspondence with major and repeating earthquakes.
Diagnostic Analysis and Prediction: From subsequent earthquakes, it is possible to identify mismatched parts of the model, and for real-time prediction, the spatial and temporal performance of online predictions is evaluated.
Challenges and Solutions: Data heterogeneity poses issues when including small earthquakes. To address historical data limitations and gaps after major quakes, we propose a new detection model combined with the HIST-ETAS model to accurately estimate activity.
Attached Figure: Serial Earthquakes in the Showa Period Nankai Trough
Note 1: The aftershock activity of the Tonankai Earthquake shifted to the Chita Peninsula (called foreshocks), leading to the Mikawa Earthquake. Note 2: A series of earthquakes were observed at the southern tip of the Kii Peninsula (within the black rectangle) from the Tonankai Earthquake to the Nankai Earthquake.
The video can be viewedhere (accessible on PC or smartphone). https://www.youtube.com/@yosihikoogata784
Background Seismicity: The HIST-ETAS model separates and extracts the typical seismic activity, using its intensity (background seismicity rate) for predictions. The intensity derived from the Japan Meteorological Agency catalog for nationwide (M>4.95) and inland (M>3.95) regions shows good correspondence with major and repeating earthquakes.
Diagnostic Analysis and Prediction: From subsequent earthquakes, it is possible to identify mismatched parts of the model, and for real-time prediction, the spatial and temporal performance of online predictions is evaluated.
Challenges and Solutions: Data heterogeneity poses issues when including small earthquakes. To address historical data limitations and gaps after major quakes, we propose a new detection model combined with the HIST-ETAS model to accurately estimate activity.
Attached Figure: Serial Earthquakes in the Showa Period Nankai Trough
Note 1: The aftershock activity of the Tonankai Earthquake shifted to the Chita Peninsula (called foreshocks), leading to the Mikawa Earthquake. Note 2: A series of earthquakes were observed at the southern tip of the Kii Peninsula (within the black rectangle) from the Tonankai Earthquake to the Nankai Earthquake.
The video can be viewedhere (accessible on PC or smartphone). https://www.youtube.com/@yosihikoogata784