Quantifying the impact of near-source uncertainties on seismic moment tensor inference

The estimation of seismic moment tensors for shallow sources is known to be highly sensitive to near-source conditions, including surface topography and shallow geological structure, which are often poorly constrained. In this contribution, we present a sensitivity analysis as part of a uncertainty quantification effort, aimed at assessing how uncertainties in near-source parameters propagate into uncertainties in inferred seismic moment tensors.
The study focuses on the Degelen Mountains dataset, which provides a well-documented setting with shallow explosive sources and complex near-surface geology.

Our methodology relies on high-fidelity 3D forward simulations of seismic wave propagation using the spectral element method, allowing accurate modeling of topographic effects and strong near-surface heterogeneities. Uncertainties in geological properties in the immediate vicinity of the source are represented using perturbations from stochastic parameterizations based on a correlation length setting, enabling the generation of physically consistent
realizations of the near-source medium. For each realization, synthetic waveforms are computed and used to perform moment tensor inversions, from which the variability of source parameters is quantified.

This framework allows us to systematically explore the sensitivity of moment tensor solutions to localized uncertainties and to identify the dominant contributors to source-related ambiguity and provide new insights into the robustness and limitations of moment tensor inversion for shallow seismic sources in complex environments.