Near-Field to Far-Field Uncertainty Propagation and Quantification of Ground Motions Generated by the Source Physics Experiments (SPE)

LLNL has developed a framework for uncertainty propagation and quantification using HPC numerical codes to simulate end-to-end, from source to receivers, the ground motions observed during the Source Physics Experiments (SPE) conducted in fractured granitic rock at the Nevada National Security Site (NNSS). SPE includes six underground chemical explosions designed with different yields initiated at different depths. To date we have successfully applied this framework to explain the near-field shear motions observed in the vicinity of SPE3 thru SPE5. However, systematic uncertainty propagation to the far-field seismic receiver has not been addressed yet. In the current study, we used a coupling between the non-linear inelastic hydrodynamic regime in the near-field and the seismic elastic regime in the far-field to conduct the analysis. Several realizations of the stochastic discrete fracture network were generated conditional to the observed sparse data. These realizations were then used to calculate the ground motions generated from the SPE shots up to the elastic radius. The latter serves as the handshake interface for the far-field simulations. By creating several realizations of near-field responses one can embed those sources into the far-field elastic wave code and further the uncertainty propagation to the receivers. We will present a full assessment from end-to-end for the near- and far-field measurements. Separate analyses of the effect of the different conceptual geological models are also carried over using a nested Monte Carlo scheme. We compare the observed frequency content at several gages with the simulated ones. We conclude that both regions experience different sampling of frequencies: small features are relevant to near-field simulations while larger feature are more dominant at the far-field. We finally rank the primary sensitive parameters for both regions to drive and refine the field characterization data collection.