Numerical Experiments of Moment Tensor Inversion with Rotational Ground Motions

Hua SiBo,Zhang Yong

DOI: https://doi.org/10.6038/cjg2022p0668

2022-01-01

Abstract:Rotational ground motions can provide spatial gradient information lacking in traditional translational ground motions, and thus can better constrain the source parameters in theory. In this study, we carry out a series of numerical experiments on moment tensor inversions with rotational motions. Firstly, we analyze the complementarity of radiation patterns of rotational and translational motions in a homogeneous infinite space, and infer that the rotational motions may provide complimentary constraints on the result if used in moment tensor inversion. Then we conduct moment tensor inversions with synthetic translational and rotational seismograms calculated with layered velocity model. We evaluate numerically the effect of using rotational components in moment tensor inversion according to the comparison of inverted moment tensor, difference angle Delta phi and seismic moment. We find that in proper condition, jointly inverting translational and rotational ground motions can significantly improve the reliability of the inversion results, and decrease the difference angle Delta phi by more than 30%, and more than 70% in some some some extreme cases. If we keep the total number of records unchanged, which means replacing translational stations with rotational components in other stations, the inversion result of moment tensor elements will be improved, while others may be worsened. Overall, adding rotational components contribute more than adding translational components of the same quantity in most station distribution. Moreover, rotational components are more sensitive to the depth at low frequency, thus they are able to provide stronger constraints on the centroid depth. In addition, we try other experiment conditions, and find the results can be further improved by using low-frequency signals close to the epicenter with low signal-to-noise ratios and accurate velocity models. On the whole, although the use of rotational components requires certain conditions, it's feasible to use them in moment tensor inversion because their spatial gradient information can provide a good supplement to the translational components. They are helpful to improve our earthquake monitoring ability in areas with sparse or dense networks.