On Using a Microearthquake Recognition System for an Early Warning System at Cotopaxi Volcano

Volcanic activity has been increasing throughout the world, posing a significant threat to populations in the event of eruptions. Ecuador, which hosts several active volcanoes, requires robust methods for identifying potential eruptions and issuing reliable alerts to protect lives and minimize damages. This paper presents the development of an automatic microseism recognition system integrated with the Early Warning Broadcast System (EWBS). Using models such as k-Nearest Neighbors, Support Vector Machine, and Decision Trees, along with frequency-based features extracted from seismic data provided by the Instituto Geofísico de la Escuela Politécnica Nacional, the recognition system aims to accurately detect and classify microeartquakes associated with the Cotopaxi volcano during 2012. During the detection stage, the system achieves an impressive Balanced Error Rate (BER) of 0.01, indicating its effectiveness in identifying events. In the subsequent classification stage, the system achieves a BER of 0.11, demonstrating its ability to classify events accurately. The classifiers were further evaluated using 82 microearthquakes, comprising 41 LP events and 41 VT events, resulting in an accuracy of 85% and a BER of 0.15. In addition, a larger data set of 563 earthquakes, consisting of 522 LP events and 41 VT events, was used to assess the performance of the classifiers. The results showed a 7% increase in accuracy compared to the previous test, demonstrating improved performance. However, 11 earthquakes were still misclassified. Integration of these classifiers with a voting system improves their performance. The selected set of 50 features plays a crucial role in achieving accurate results. The recognition system seamlessly interfaces with the EWBS, ensuring a 30 s delay before launching an early warning. This delay provides valuable time for preparation and response measures. In conclusion, the developed microearthquake recognition system, combined with the EWBS, demonstrates its effectiveness in detecting and classifying events, thereby enhancing the ability to issue timely and reliable alerts for volcanic activity. The findings contribute to improved volcano monitoring and risk mitigation strategies.