Modeling Travel-Time Correlations Based on Sensitivity Kernels and Correlated Velocity Anomalies
Abstract
A major contributor to the error in regional and teleseismic event locations is travel-time prediction error, attributable to velocity anomalies in the real Earth that are not represented in the reference Earth model an event locator uses for travel-time calculation. The prediction errors at two stations will, in general, be correlated depending on the spatial coherency of the velocity anomalies compared to the distance between the stations. The motivation for this project is the observation by previous workers that large event location errors can be induced when prediction error correlations are ignored and the observing network of stations is far from uniform. Recently, event location algorithms have been generalized to accept a non-diagonal covariance matrix for data errors as a mechanism for accommodating correlated errors in travel-time predictions. Our project is addressing what to input for the covariance matrix, based on considerations of seismic wave propagation and the statistical characterization of the Earth's velocity heterogeneity. Specifically, we are developing numerical algorithms that generate a travel-time covariance matrix for a network of stations, as a function of the event location, by integrating a velocity covariance function, as defined in geostatistical analysis, with the travel-time sensitivity kernels appropriate for the event-station paths. This paper describes two alternative techniques for performing this calculation, which are compared in a numerical example. A second example of our approach investigates the variation of travel-time variances and correlations over regional and near-teleseismic distances, computed using ray sensitivities for a 1D reference Earth model. This example demonstrates a strong dependence on ray parameter that agrees qualitatively with empirical observations.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2007
- Accession Number
- ADA519772
Entities
People
- Stephen C. Myers
- William L. Rodi
Organizations
- Massachusetts Institute of Technology