Seismic Classification

Earthquake classification is an important tool for understanding the similarities and differences between earthquakes. By putting earthquakes into classes, we are able to study patterns and features that apply to more than just one earthquake, which enables the development of better seismic models. However, past earthquake classification schemes have largely been rough guides that fail to group earthquakes based on the most commonly used metrics. Here we develop a new earthquake classification scheme for volcanic earthquakes wherein each class is differentiated by the most commonly used seismic properties. In order to accomplish this, we used a deep learning algorithm to find patterns in the seismic data of earthquakes with similar seismic properties. Past work has demonstrated that chaotic models can be used to model low amplitude earthquakes, so we then used another algorithm to fit a chaotic model to the patterns found with deep learning and we defined our class boundaries from the boundaries of these models. This classification scheme can now be used to concretely classify earthquakes based on metrics geophysicists often use.