Time series clustering study of Sumatran earthquakes.

Abstract

Earthquakes cause tremendous destructions every year in Sumatra. In order to minimize the loss, researchers are trying very hard to predict earthquakes. However, there is no satisfactory way of prediction currently. Good prediction follows deep understanding. The purpose of this project is to gain deeper insight of the correlation between earthquakes and relative movements of GPS stations. We hope to find phenomenon that can be used as a precursor. Time series clustering method is applied. By clustering the Sumatra GPS stations at each period according to their relative movements and monitoring the changes of clusters with time, we find many evidences that suggest the correlation between the earthquakes and the fusion or fission behaviors of clusters. We also find that the clusters behave differently at different phases of a huge earthquake.