Long-term evolution of continuous seismic signals at Piton de la Fournaise volcano inferred from the network covariance matrix.

Continuous seismic signals recorded in the vicinity of active volcanoes are composed of seismic waves generated by a variety of internal and environmental sources and propagating through different parts of the plumbing system. This implies that these signals are very sensitive to the state of the plumbing system. A change in the volcanic activity affects the properties of the seismo-volcanic sources while a change in the plumbing structure affects the media through which the seismic waves propagate. Network-based analysis of continuous seismic records has been developed to incorporate information from multiple stations simultaneously. Here we use an approach based on the network covariance matrix that combines an ensemble of inter-station cross-correlations. We compute the width of the eigenvalue distribution of this matrix at a given frequency in a moving time window, resulting in a compact time-frequency representation of continuously recorded seismic wavefield.

 We apply this analysis to ten years (2013-2023) of continuous seismic data from the Piton de la Fournaise volcano located in la Réunion, France. The resulting spectral width distributions indicate that continuous signals are characterized by multiple narrow spectral peaks, which are observed during co-eruptive tremors as well as during periods without visible volcanic activity. We propose a normalization process to enhance these peaks in both the frequency and time domains. We observe numerous spectral peaks in the 1-3 Hz frequency band that remain nearly constant for extended periods (weeks to months). We observe a distinct difference in the spectral peak distribution between co-eruptive and quiet periods, as well as significant variations during long-standing eruptions. Locations of sources of the co-eruptive signals correlate well with the eruption sites. The inter-eruptive signals seem to originate from a combination of environmental and weak internal sources, and changes in their spectral properties might reflect the medium changes after major eruptions.