End-to-end LSTM based estimation of volcano event epicenter localization

Locating sources of volcano-seismic event is very relevant to monitor and comprehend volcanic processes. Ordinary estimation source seismic events based on phase picking. The most accurate procedure selection the visual inspection records by experts, who employ local characteristics for detection comparison with observed signals from other stations. This activity highly time demanding, which in turn a strong motivation automatize epicenter process. However, automatic picking volcano inaccurate because short distances between epicenters seismograph In this paper, an end-to-end LSTM (Long-Short Term Memory) scheme proposed address problem localization without any priori model relating estimation. was chosen due its capability capture dynamics varying signals, remove or add information within memory cell state long-term dependencies. A brief insight into also discussed here justify use neural network. results presented paper show that architecture provided success rate, i.e., error smaller than 1.0 km, equal 48.5%, dramatically superior one delivered Moreover, method gave rate (18%) higher CNN (Convolutional Neural Network). suggest approach can be applied geophysics problems.