Mining Extreme Patterns of Seismic Signals in China Based on Heavy-Tailed Statistics

摘要

Strong earthquakes are the most violent natural hazards, which have caused huge losses of lives and property to human beings in the last decade. Mining time patterns of seismic events, especially extreme ones, within a specific time window and region, is of great importance to policy making and disaster reduction. In this paper, we apply heavy-tailed statistics in extreme value theory to mining the temporal pattern of seismic events in China and build a computational model. We show that earthquakes of magnitude above Ms. 6.0 in China follow the generalized Pareto distribution. We use the proposed model to simulate the possible risk of extreme seismic magnitude in future time windows. Comparison with real-world records demonstrate the validity of our model.