Machine Learning for Volcano-Seismic Signals: Challenges and Perspectives

International audienceEnvironmental monitoring is a topic of increasing interest, especially concerning the matter of natural hazards prediction. Regarding volcanic unrest, effective methodologies along with innovative and operational tools are needed to monitor, mitigate and prevent risks related to volcanic hazards. In general, the current approaches for volcanoes monitoring are mainly based on the manual analysis of various parameters, including gas leaps, deformations measurements and seismic signals analysis. However, due to the large amount of data acquired by in situ sensors for long term monitoring, manual inspection is no longer a viable option. As in many Big Data situations, classic Machine Learning approaches are now considered to automatize the analysis of years of recorded signals, thereby enabling monitoring at a larger scale. This paper focuses on integrated and operational tools dedicated to the automatic analysis of volcano-seismic signals. Namely we review (i) tools for the optimal representation of volcano-seismic signals (feature space) and the available methods for volcano-seismic events (ii) detection and (iii) classification. We then propose an architecture for the automatic classification of volcano-seismic events. Our prediction system is tested on 6 years of recordings containing 109434 volcano-seismic events acquired from Ubinas volcano (the most active volcano in PerúPer´Perú). Our new proposed model is build using supervised machine learning algorithms (Support Vector Machine) and reaches 92.2% of correct classification over six classes. This prediction model is then used to fully analyze the 6 years of recorded signals