Yield Estimation from Waveform Inversion of Regional Infrasound Signals

We investigate a full-waveform inversion technique for infrasound signals and its application to explosion yield estimation at regional ranges (15 - 250 km distance). Yields of explosions are often determined by infrasound signals based on empirical models, which provide relationships between the explosion energy and the metrics of infrasound signals (e.g., amplitude, duration, and impulse per unit area). Meteorological parameters (e.g., temperature, pressure, and wind) have significant impacts on regional infrasound propagation and, therefore, the use of appropriate model accounting for meteorological effects is critical to the accuracy of yield estimation. However, existing empirical models often do not include a large variety of meteorological scenarios, and an empirical model determined in a certain meteorological scenario may not be valid in different weather conditions. In this study, we propose a method based on numerical models generated by infrasound propagation simulations, which takes into account diverse meteorological conditions. A full 3-D finite difference method is used for the simulations providing Green's functions for infrasound observations. Full-waveform inversions (Kim and Rodgers, 2016) are performed using the numerical Green's functions to obtain acoustic source time histories of events, and yields of explosions are determined from the characteristics of the resultant equivalent acoustic sources (e.g., amplitude, duration, impulse, and acoustic energy). The method is applied to controlled ground-truth infrasound sources for verification, and theoretical uncertainty of the method is evaluated statistically.