A Study of Earthquake Seasonality in Japan Through Network Analysis

Abstract

Earthquakes are some of the most mysterious events that occur on planet Earth. Their complex nature makes them difficult to analyze and fully understand. In an attempt to look at earthquakes and the many factors that contribute to them, it is essential to break down the multitude of questions surrounding the natural occurrence into much smaller and focused analyses. In order to better predict earthquakes, seismologists and data scientists look for patterns in the data sets. This thesis looks at earthquake seasonality through node importance, a measure of how important nodes are relative to one another. By framing earthquake data as a network, with regions as nodes and successive earthquakes as edges/links, PageRank analysis, closeness centrality and betweenness centrality are used to find any major differences in node importance measures across all four seasons in Japan. Looking at the results as a directed graph and geographically as well, no major differences are found in node importance across the algorithms and across seasons in Japan.