Petrophysical Inversions from Seismic Images detect and characterize Melt under Toba Caldera (Indonesia).

The crustal feeding systems under volcanoes like Toba caldera in Sumatra remain largely unknown due to the lack of existing seismic arrays and consistent effort to apply advanced imaging techniques. Even if they were obtained, the interpretation of seismic anomalies in terms of melt content and geochemical differentiation would remain largely speculative without a framework that allows the connection between seismic attributes and petrophysical properties.

We collected data from existing seismic arrays across Sumatra and applied ambient noise and earthquake tomography techniques. We used thousands of dispersion curves spanning both the shallow crust and the upper mantle using SeisLib¹, the first open-access Python package for surface-wave tomography. We inverted the dispersion curves for phase-velocity maps at different periods, using a linearized-inversion algorithm based on the ray theory with a roughness damping constraint and an adaptive parameterization. Our results show low-velocity and high-attenuation sill-like structures under most of the northern and central portions of Toba calderas. We used the shear-wave velocity model as data and additional volcanological and geophysical data as constraints for a Bayesian inversion of magmatic composition and melt content under Toba. The forward model is provided by the MAGEMin³ code, which uses a Gibbs energy-minimization solver coupled with geological, geophysical, and volcanological information to identify portions of the crust where mafic sills are located. The Bayesian inversion quantifies from the melt content within the sill-like structures to the characteristics of the caprock overlaying them. It confirms the existence of a deep extended mafic sill reaching depths up to 12 km, currently disconnected from similar pockets of melt underneath the Toba lake stalling at a similar depth.

By coupling seismic and thermodynamical modeling, we invert for (petro)physically-constrained quantitative images of the current state of a volcano. These images provide a cornerstone for a temporal description of volcanological responses based on physical modeling.

• 1) Magrini, Fabrizio, et al. "Surface-wave tomography using SeisLib: a Python package for multi-scale seismic imaging" Geophysical Journal International (2022). ggac236, https://doi.org/10.1093/gji/ggac236.
• 2) Riel, Nicolas, et al. "MAGEMin, an efficient Gibbs energy minimizer: Application to igneous systems" Geochemistry, Geophysics, Geosystems (2022).