Estimating near-surface explosion yield and its uncertainty using regional seismic envelopes

To improve the estimation of near-surface explosive yield at regional distances for the nuclear forensics application (e.g., Stone, 2016), we are incorporating improved predictions of seismic waveform envelopes and a calibrated propagation model that includes uncertainty. Our approach computes waveform envelopes using a new hybrid computational technique that includes both multiple forward- and wide-angle-scattering effects (e.g., Saito et al., 2003). The new modeling technique improves predicted envelope fits over the entire P-waveform and accounts for direct-wave and early coda complexity. We use an extensive training dataset of earthquakes in the United States to calibrate the waveform envelopes as well as the propagation model for laterally-varying scattering/absorbing structures. The new hybrid envelope with the transport model more accurately predicts P-wave energy propagation for the continental United States, and it also incorporates rigorous uncertainty estimates. The predicted P-wave energy is then used in a Bayesian inference scheme to estimate a maximum-likelihood solution for near-surface explosion yield and its uncertainty. We will demonstrate the new technique with regional seismic records from near-surface explosions in the United States.

References

Saito, T., H. Sato, M. Fehler, and M. Ohtake (2003), Simulating the envelope of scalar waves in 2D random media having power-law spectra of velocity fluctuation, Bull. Seismol. Soc. Am., 93, 240-252.

Stone, R. (2016), Surprise nuclear strike? Here's how we'll figure out who did it. Science, doi:10.1126/science.aaf4194