Self-supervised learning for the exploration of continuous seismic records at the Fani Maoré submarine volcano (Mayotte)

Continuous seismological observations provide valuable information to deepen our understanding of processes occurring in both aerial and submarine volcanoes. However, the wealth of the seismicity recorded near volcanoes makes exhaustive exploration of these seismological chronicles very complex and time-consuming. In this study, we present a systematic analysis of two months of seismological records using a self-supervised learning (SSL) approach for the unsupervised clustering of continuous seismic data acquired by ocean bottom seismometers deployed in the vicinity of the Fani Maoré volcano (Mayotte). The proposed clustering process allows the identification of individual seismic events, seismic crisis and tremors that would be challenging to observe using conventional approaches. We show that our approach detects and classifies both known and new events, including two eruptive sequences previously unknown. We also demonstrate the potential of self-supervised methods for the analysis of seismological records, providing a synoptic view and facilitating the discovery of insightful yet rare events. This approach has numerous applications in exploring various seismological datasets, simplifying analysis while making it more comprehensive.