Developing Path-Dependent Uncertainty Estimates for use with the Regional Seismic Travel Time (RSTT) Model

Friday, 18 December 2015: 08:15
305 (Moscone South)
Michael L Begnaud1, Dale N Anderson1, W. Scott Phillips1, Sanford Ballard2 and Steve Myers3, (1)Los Alamos National Laboratory, Los Alamos, NM, United States, (2)Sandia National Laboratories, Albuquerque, NM, United States, (3)Lawrence Livermore National Laboratory, Livermore, CA, United States
The Regional Seismic Travel Time (RSTT) tomography model has been developed to improve travel time predictions for regional phases (Pn, Sn, Pg, Lg) in order to increase seismic location accuracy. The RSTT model is specifically designed to exploit regional phases for location, especially when combined with teleseismic arrivals. The latest RSTT model (version 201404) has been released (http://www.sandia.gov/rstt). Travel time uncertainty estimates for RSTT are determined using one-dimensional (1D), distance-dependent error models, that have the benefit of being very fast to use in standard location algorithms, but do not account for path-dependent variations in error, and structural inadequacy of the RSTTT model (e.g., model error). Although global in extent, the RSTT tomography model is only defined in areas where data exist. A simple 1D error model does not accurately model areas where RSTT has not been calibrated. We are developing and investigating a new covariance matrix for RSTT phase arrivals by mathematically deriving this multivariate error model directly from a unified model of RSTT embedded into a statistical random effects model that captures distance, path and model error effects. An initial method developed is a two-dimensional path-distributed method using residuals. Other methods include a complete random-effects model and the calculation of the full model covariance matrix from the RSTT tomographic inversion. The goals for any RSTT uncertainty method are for it to be both readily useful for the standard RSTT user as well as improve travel time uncertainty estimates for location.