[S06-1] Advancement in methodologies for CTBT monitoring

[S06-1-04] Model ensembles for estimation of seismic travel time and event location uncertainty

^○Stephen Myers, Nathan Simmons (Lawrence Livermore National Laboratory, Livermore, California, USA)

We explore the use of model ensembles to characterize the uncertainty of seismic travel time predictions. Our emphasis is on 3D earth models and we include a standard, radially symmetric (1D) models for comparison. Traditional parameterization of travel time prediction uncertainty – as a function of event-station distance – provides a representative global average, but uncertainty characterization can be unrepresentative when errors throughout a large geographic region deviate significantly from the average. Recent efforts to characterize path-specific travel time uncertainty have made great strides, but ad hoc multiplication factors must be applied to uncertainty estimates for them to be reconciled with observed error. The mismatch between estimated uncertainty and observed error is likely due to epistemic errors, for which model-error covariance is typically unknown. The model ensemble approach leverages the efforts of many research groups that have rendered images of earth structure. To the extent that researchers use differing data sets, model parameterization, methods of travel-time prediction, or imaging, predictions based on an ensemble of models will improve robustness of travel time uncertainty. In our initial effort, we use a set of seismic models to compute P-wave travel times for a global set of events with well-characterized location uncertainty. Agreement and divergence of travel time predictions for each model globally, regionally, and for each path are examined to study the variability of model predictions. The test events are relocated and epicenter uncertainties are estimated using each model and using estimates of uncertainty based on an ensemble calculation. The consistency of location errors and uncertainties are used to assess the model ensemble approach. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.