Toward Improving Point-Source Moment-Tensor Inference by Incorporating 1D Earth Model's Uncertainty: Implications for the Long Valley Caldera Earthquakes

To infer seismic moment tensor (MT) of moderate earthquakes at regional scales, seismologists typically simulate waveforms using available structural models of the Earth to match the observed seismograms. This procedure is known as waveform seismic MT inversion. However, the seismic data are noisy, and the Earth model is inevitably different from the actual structure of the Earth; hence, there is a discrepancy between the predicted and the observed waveforms. This discrepancy arises from the noise in the data and imperfections in theoretical predictions, stemming most significantly from the Earth model. This study introduces structural uncertainty, estimated empirically, and referred to as theory uncertainty, as part of the combined covariance matrix. In the synthetic setting, we first show through a series of synthetic experiments that the structural uncertainty plays a critical role in retrieving MT solutions, especially for short-period waveforms. The method is then benchmarked against the waveforms of non-double-couple earthquakes in Long Valley Caldera, California. We confirm the highly isotropic nature of the source in a pilot event but find a non-negligible CLVD component that was overlooked in past studies ignoring the uncertainty in Earth model. Thus, careful consideration of the Earth model's uncertainty as part of the MT inversion schemes will be necessary for future applications to better understand the complicated physics of earthquake sources. Plain Language Summary The representation of a seismic source, which can be an underground faulting, a volcano-unrest event, or a manmade explosion, is often through an equivalent force system placed at a point, known as a point-source moment tensor (MT). Its inference from the observed seismic waveforms is referred to as an MT inverse problem. This study is at the forefront of solving this inverse problem by accounting for two primary sources of unavoidable uncertainties: (a) structural uncertainty related to the imperfection of the Earth model representing the true Earth, and (b) data noise stemming from the observed time series of ground motion. Although the former outweighs the latter in practice, it is often ignored. Here, we show that an appropriate account of both uncertainties is critical in yielding reliable MT solutions. It allows for the meaningful interpretation of the nature of seismic sources, including nondouble couple events and clandestine nuclear explosions. We illustrate this both through comprehensive synthetic experiments and an in-depth analysis of a volcanic event in the Long Valley Caldera, CA, for which a significant nondouble-couple component is now revealed.

Automated summary

This study introduces and highlights the importance of structural uncertainty, along with data noise, in inferring seismic moment tensor (MT). By incorporating both uncertainties, the study shows that it is critical in yielding reliable MT solutions and interpreting the nature of seismic sources accurately.