Source Inversions of Volcano Infrasound: Mass Outflux and Force System for Transient Explosive Eruptions

Abstract:

Sources of volcano infrasound associated with explosive eruptions are typically modeled assuming an acoustic monopole and/or dipole. While the monopole represents the mass outflux of erupted materials, the dipole represents a force system acting in the source region during eruptions. Therefore, appropriate acoustic source inversions of volcano infrasound data can provide estimates of eruption parameters which are critical to understanding eruption dynamics. Reliability of the source parameters is dominantly controlled by the accuracy of the acoustic Green's functions between the source and receiver positions. Conventional source inversions of volcano infrasound, however, were typically performed using a simplified Green's function obtained in a free space or half space. This may result in intolerable errors in the source parameters, especially when the infrasound waveforms are significantly distorted by volcano topography and/or local atmospheric variability (i.e., layered velocity structure or wind). In this study we present a full waveform inversion technique for volcano infrasound using numerical Green's functions. A full 3-D Finite-Difference Time-Domain (FDTD) method accelerated with GPU is used to compute accurate Green's functions taking into account volcano topography and local atmospheric conditions. The presented method is applied to data recorded at Sakurajima volcano (Japan) and Tungurahua volcano (Ecuador), which provide a large volume of high-quality data recorded by azimuthally well-distributed stations within 2 -- 6 km distance of the volcanoes. We analyze infrasound signals associated with explosive eruptions exhibiting 1) distinct explosion waveforms followed by gas discharges and 2) strong anisotropic radiation patterns, which can be caused by either source directivity or topographic barriers/reflections. Here the role of topography in controlling the infrasound radiation is investigated through numerical modeling, and then the observed infrasound is inverted for the source parameters by using the modeled Green's functions. Finally, we investigate explosive source mechanisms for different types of eruptions considering the obtained monopole and dipole components.