Dynamic Neural Networks for Classification of Volcanic Earthquakes

Volcanic earthquakes are often classified into several broad categories, including volcano-tectonic, long-period, hybrid, and rockfall events. Manual classification is time-consuming and subjective, and a reliable method to automatically classify these earthquakes in real-time is needed. Several approaches have been proposed for this goal, including the use of hidden Markov models and artificial neural networks (ANNs). Here we present preliminary results of an ANN-based earthquake classification system. Numerical parameters such as mean amplitude and frequency content are first calculated from seismic waveforms, then fed into a dynamic neural network setup for training and testing. The results are compared to those obtained using principal component analysis, a traditional statistical approach.