Insights into Yellowstone’s magmatic system from 3D seismic waveform modeling and inversion

Abstract:

The Yellowstone hotspot, which has hosted three large silicic caldera forming eruptions in the last ~2 Myr, is thought to be fueled by an extensive magmatic system that may span the crust and uppermost mantle. Our understanding of Yellowstone’s magmatic system has been informed in large part by imaging wavespeed variations in the subsurface using seismic tomography. In recent years, the application of ambient noise techniques has greatly improved the spatial coverage of surface wave dispersion datasets and has enabled new insights into the structure and melt organization of the magmatic system. However, it is poorly understood whether or not traditional tomography techniques based on inverting surface wave dispersion observations can reliably capture the extreme wavespeed variations expected to accompany significant partial melt fractions in the crust. Here, we perform a set of synthetic resolution tests based on 3D spectral element simulations of seismic waves propagating through representative models of Yellowstone’s magmatic system. Our tests mimic the source-receiver geometry made possible by the deployment of broadband seismic networks in the Yellowstone region over the last ~20 years. We explore scenarios in which the maximum shear velocity reduction in the model ranges between -10% to -66%, which encompasses the range of end-members informed by either seismic tomography or wave scattering studies. The magmatic anomaly is localized below the Yellowstone caldera between depths of 5 - 15 km, consistent with previous seismic imaging. We find that even for idealized imaging geometries, recovered phase velocity anomalies at periods between 6 - 25 s are strongly diminished, implying that estimates of the melt fraction based on traditional surface wave tomography may be biased downwards. To overcome the limitations of traditional techniques, future efforts should focus on adjoint inversions of phase delays, which incorporate more accurate sensitivity kernels based on realistic wave propagation physics.