Abstract

Volcano monitoring is very important in many countries. Chile is located in the Pacific fire belt, so the country has hundreds of volcanoes along its territory that may affect the economic and social activity. The main variable to monitor is volcano seismicity, as the recognition of seismic patterns originated inside the volcano gives important information about its activity. The great number of volcanos and the huge amount of seismic records makes necessary to automate some steps of this process. This papers presents a hierarchical structure based on Support Vector Machines (SVM) to classify four classes of seismic events recorded by the seismic stations. The implementation of the structure was performed using the database of the Llaima volcano and the classifiers were tested in another volcano, the Nevados de Chillán Volcanic Complex. The results show a good performance with a mean accuracy of 89.7% for the Llaima volcano and 84.9% for the Chillán volcano, with no retraining process for this new seismic dataset