Inversion of Infrasound Time Series for Seismoacoustic Source Parameters Produced by a Buried Chemical Explosion at the Source Physics Experiment Phase II: Dry Alluvium Geology

Abstract

ABSTRACT We invert infrasonic timeseries produced by a single buried chemical explosion detonated as part of the Source Physics Experiment phase II Dry Alluvium Geology (DAG). The key to our inversion is that we use a three dimensional, fully coupled, linear, elastic-to-acoustic forward model to predict the recorded acoustic wavefield. We assume that the fourth buried chemical explosion in the series, DAG-4, produces an atmospheric acoustic wavefield due to a combination of direct elastic-to-acoustic coupling at the air–earth interface and ground upheaval, or spall, at the Earth’s surface. In our linear model we approximate these two phenomena with distinct source terms separated in space and time: (1) a buried point source consisting of six source time functions, each corresponding to a single component of a generalized moment tensor and (2) a vertically directed, time-variable force applied to the Earth’s surface. Inverting the infrasonic data with this linear forward model results in estimated source parameters that accurately predict the observed infrasound signals. To test our estimated buried source, we use the Rayleigh integral (RI) to model the infrasound signal from the measured and predicted acceleration surfaces. First, we simulate the infrasound signal waveforms using vertical accelerometer data and find that these closely match the observed infrasound. Next, using the estimated buried seismic source model without spall, we estimate the acceleration of the Earth’s surface using a linear approximation. When using the predicted surface acceleration in the RI, we simulate infrasound signals that reproduce pulse shape characteristics but not the amplitude of the observed infrasound, indicating our linear model does not completely account for nonlinear spall effects. Based on these results, we argue that for the scale of this experiment, purely linear models can estimate seismoacoustic source attributes and predict the far-field infrasonic signal, but the primary contribution to the infrasonic signal is the spall source.