Multi-scale reasonable attenuation tomography analysis (MuRAT): An imaging algorithm designed for volcanic regions

• Coda-normalization is a powerful method for measuring average seismic attenuation.
• MuRAT is designed to image seismic attenuation in volcanic and geothermal areas.
• The multi-step code increases stability in the presence of high heterogeneity.
• Heterogeneous scattering affects attenuation tomography in/under volcanic cones.
• MuRAT provides information on the feasibility of attenuation tomography in volcanoes.

The attenuation of body-wave amplitudes with propagation distance can be used to provide detailed tomographic images of seismic interfaces, fluid reservoirs, and melt batches in the crust. The high sensitivity of body-wave energies to high-scattering structures becomes an obstacle when we try to apply attenuation tomography to small-scale volcanic media, where we must take into account the complexities induced by strong heterogeneous scattering, topography, and uncertain source modeling in the recorded wave-fields. The MuRAT code uses a source- and site-independent coda-normalization method to obtain frequency-dependent measurements of P-to-coda and S-to-coda energy ratios. The code inverts these data for both the geometrical spreading factor and the spatially-dependent quality factors (Q), providing additional attenuation information in the regions where velocity tomography is available. The high sensitivity of coda-waves to highly heterogeneous structures highlights zones of anomalous scattering, which may corrupt amplitude-dependent attenuation measurements, and where basal assumptions of linear optics may go unfulfilled. A multi-step tomographic inversion increases the stability of the results obtained in regions of high heterogeneity (e.g., the volcanic edifice) by the inclusion of data corresponding to either sources or stations located in regions of lower heterogeneity. On the other hand, a mere increase in the number of rays entirely contained in the heterogeneous structures affects both the stability and the effective resolution of the results. We apply the code to two small waveform datasets recorded at an active (Mount St. Helens) and at a quiescent (Mount Vesuvius) volcano. The results show that the seismicity located inside or under the volcanic edifice produces an increase of the low-frequency energy ratios with travel time in both areas. In our interpretation, the anomalous concentration of energy which affects any waveform recorded on the cone, produced inside the volcanic edifice or in the feeding system of the volcano is due to seismic source- or medium-dependent resonance. The results also provide spatial and frequency limits on the feasibility of attenuation tomography in these two regions with larger datasets.