Hannes Vasyura-Bathke, Jan Dettmer, Rishabh Dutta, Paul Martin Mai, Sigurjón Jónsson

Earthquake-source parameters can be estimated from seismic waveforms. Since these data indirectly observe the deformation process, parameters of a physical model that quantifies the deformation process are inferred through the inverse problem; which is under-determined. This requires several assumptions to be made about Earth structure and other aspects that affect the source parameter estimation. These assumptions primarily include a simplified seismic velocity model of the Earth waveform and noise models. The specific model choices affect data residuals and can lead to biased source parameter estimations and unrealistic assessment of the associated source-parameter uncertainties. While data errors are routinely included in parameter estimation for full centroid moment tensors, less attention has been paid to theory errors related to velocity model uncertainties and how these affect the resulting moment-tensor uncertainties. Here, we study non-linear full moment tensors with several simulated data sets and demonstrate that subsurface structure uncertainties can profoundly affect parameter estimation and that their inclusion leads to more realistic parameter uncertainty quantification. We present a solution to include model errors by estimating non-stationary (non-Toeplitz) error covariance matrices that lead to appropriate source-parameter estimates and uncertainties.
Finally, we demonstrate the influence of these noise parameterisations on real regional seismic data of the ML 4.4, 13 June 2015 Fox Creek event, Canada.
Including uncertainties in Earth-structure resulted in robust source parameter estimates in case the structure was poorly known.