Model-Based Signal Processing Approach to Nuclear Explosion Monitoring

摘要

This report describes research performed under Laboratory Research and Development Project 05-ERD-019, entitled A New Capability for Regional High-Frequency Seismic Wave Simulation in Realistic Three-Dimensional Earth Models to Improve Nuclear Explosion Monitoring. A more appropriate title for this project is A Model-Based Signal Processing Approach to Nuclear Explosion Monitoring. This project supported research for a radically new approach to nuclear explosion monitoring as well as allowed the development new capabilities in computational seismology that can contribute to NNSA/NA-22 Programs. Current methods for seismic monitoring reduce seismograms to a small number of derived measurements from observed signals, such as the arrival times and amplitudes of waves with specific propagation paths through the Earth (i.e. phases). These measurements are interpreted in terms of deterministic one-dimensional (1D) Earth models, possibly with empirical or model-based corrections for three-dimensional (3D) structure. These models are often inaccurate due to our lack of knowledge of detailed sub-surface structure on a wide range of length-scales. Our lack of knowledge of structure generally becomes more important as the frequency of interest increases, inhibiting our ability to model high-frequency (> 1 Hz) signals and requiring the use inherently limited empirical calibration strategies to monitor small events at regional distances (