Research

Source mechanisms of earthquakes can be determined from the waveforms or spectra of seismic waves, and are described by moment tensors (MTs) which provide information beyond slip on a planar fault. The deployment of global digital seismic networks allowed development of large databases of MTs. Issues about the documentation of their inversion procedure (Rösler et al., 2002b) leave questions about these catalogs. My research explores these questions by taking a big data approach using MT catalogs to gain insight into general properties of earthquakes.

Moment tensors can be decomposed into components representing geologic processes. The most common decomposition is into an isotropic component, a double-couple component (DC) and a compensated linear vector dipole component (CLVD). While CLVD components can reflect complex source processes for earthquakes in specific geologic environments or near-simultaneous earthquakes on nearby faults of different geometries, they may also be artifacts resulting from the inversion process used to derive moment tensors by finding the best fit to the data. Its results depend on the seismic phases inverted, the Earth model for elastic and anelastic structure assumed, noise in the data, and the number of seismic stations and their azimuthal coverage.

Uncertainties in Moment Tensors

Due to the nonlinearity of the inversion, quantifying the uncertainties is difficult. As a consequence, they are either not reported or estimated from the misfit between the synthetic and observed waveforms. However, these uncertainties only account for errors in the data, and ignore modeling errors. To assess the uncertainties in moment tensors, I examined the differences between 5,000 moment tensors in the GCMT and USGS catalogs and found that the differences are typically an order of magnitude larger than the reported errors (Rösler et al., 2021). DC components and hence fault angles of earthquakes are determined with greater certainty than CLVD components. CLVD components are moderately correlated between catalogs, and agree better between catalogs for larger earthquakes. This finding suggests that the NDC components of many small earthquakes are likely to be artifacts of the inversion without geological meaning.

Origin of Non-Double-Couple Components in Moment Tensors

I examined this hypothesis in another publication (Rösler and Stein, 2022a) using moment tensors of three global and four regional catalogs for 2016-2020. This provides a dataset of NDC components of 12,856 earthquakes with a large range of magnitudes in various geologic environments. The NDC components are essentially independent of magnitude, with a mean deviation from a double-couple (DC) source of around 20%. The consistency suggests that most NDC components do not reflect rupture on multiple faults, which only occurs for larger earthquakes. Furthermore, there are only small differences in mean NDC components between earthquakes with different faulting mechanisms, or in different geologic environments. This consistency suggests that most NDC components do not reflect actual source processes, which would likely cause variability. Hence it appears that for most earthquakes, especially smaller ones, the NDC components are likely to be artifacts of the inversion.

Non-Double-Couple Components as Artifacts of the Moment Tensor Inversion

Numerical experiments have confirmed this result (Rösler et al., 2022c). By generating synthetic seismograms for DC MTs for one Earth model, and inverting them using Green's functions generated for other Earth models, I showed that to match the waveforms with a different Earth model, the inversion changes the mechanism to include a substantial NDC component. The NDC components generated during the inversion are comparable to those of earthquakes reported in the GCMT catalog, which implies that these components are largely artifacts of the inversions not adequately accounting for the effects of laterally varying Earth structure.

Reliability of Non-Double-Couple Components

In order to assess when NDC components are likely to represent real source processes, I analyzed the differences between three global MT catalogs (Rösler et al., 2023). Because the catalogs use different inversion procedures, the standard deviation among the NDC components is a measure of the reliability of their determination. A decrease in standard deviation with magnitude indicates increased reliability of the NDC components for larger earthquakes. A decrease in standard deviation with size of the NDC components, however, appears to be partially a consequence of the definition of the NDC components. Through comparison with random NDC components, I found that NDC components that are likely to represent real source processes are expected to be larger than 60%. This threshold well exceeds the global average of 20%.