← Back to Sessions

Estimating Explosion Source Time Functions From Seismic Data: Are Linear Inversions Good Enough?

Session: Explosion Seismology Applications and Advances [Poster]

Type: Poster

Date: 4/19/2021

Presentation Time: 11:30 AM Pacific

Description: 

When inverting seismic data for seismic source parameters, it's common to simplify the forward model by using linear approximations. However, by their very nature, linear inversion schemes cannot account for, or predict, the effects arising from nonlinear phenomena of the source. Such phenomena may include processes like rock fragmentation, pore space crushing, and heat generation. These phenomena may give rise to spurious seismic waveforms that are not predicted by our linear models. Therefore, our motivating questions for this study are 1) how well do linear inversion schemes estimate source parameters when the data are known to be affected by nonlinear, near-source processes, and 2) how well can the linear inversions fit the this data? We perform a series of idealized numerical experiments on synthetic data and illustrate various nonlinear source-related phenomena on the seismic data as well as their effects on the inversion results. To produce the simulated data we numerically couple a nonlinear shock physics code to a linear, finite difference code. The shock physics codes simulates the nonlinear phenomenology associated with a buried chemical explosion, where the resulting seismic wavefield is coupled into the linear finite difference code. We simulate data for several different explosion-source emplacement scenarios: a fully tamped explosion and several explosions that are detonated inside air-filled cavities of various volumes and shapes. In all cases, the simulated sources are the same yield. For our tests, we found that the average P-wave amplitude is inversely proportional the volume of the air-filled cavity, which is the well-known decoupling problem. Also, we observed that hemispherical, air-filled cavities generate shear waves, which simple linear inversions do not predict. Although the data is predicted reasonably well using our linear inversion scheme for all the tests here, the quality of the data fit decreased as the volume and/or geometrical complexity of the air-filled cavities increased.

Presenting Author: Christian Poppeliers

Student Presenter: No

Authors