Error Analysis in the Joint Event Location/Seismic Calibration Inverse Problem
Abstract
This project is developing new mathematical and computational techniques for analyzing the uncertainty in seismic event locations, as induced by observational errors and errors in travel-time models: The analysis is being done in the context of the multiple-event inverse problem, in which the locations of multiple events are inferred jointly with travel-time corrections for the event-station paths. The premise is that one of the events is the target of the uncertainty analysis while the others are calibration events that serve to constrain the path corrections within some level of error. An uncertainty analysis on the coupled location/calibration inverse problem leads to the notion of a "multiple-event" confidence region on the target event location, which accounts implicitly for errors in the inferred path corrections, including the effects of uncertainty in the calibration event locations. The approach we have developed considers the nonlinearity of the forward problem and allows the use of non-Gaussian models for data errors and soft or hard constraints on the problem parameters. The project has addressed mainly the basic multiple-event location problem, wherein travel-time corrections comprise a simple time term for each station/phase combination in the data set. We have developed a numerical scheme for computing multiple-event confidence regions for this problem, based on grid search and Monte Carlo sampling techniques. Given the computationally intensive nature of the approach, some approximations were developed to reduce the computational effort. Experiments with data from the Nevada Test Site validated the approach and demonstrated the adequacy of the approximations when the location uncertainty is not too large. A modification of the approach we are investigating, which offers the possibility of additional efficiency, is the computation of multiple-event confidence regions defined in a Bayesian sense.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2006
- Accession Number
- ADA469438
Entities
People
- William L. Rodi
Organizations
- Massachusetts Institute of Technology