Relative Moment Tensor Inversion for Microseismicity: Application to Clustered Earthquakes in the Cascadia Forearc

The relative abundance of small earthquakes affords significant opportunities for improved understanding of regional seismotectonics; however, determining moment tensors for such events recorded on regional networks is complicated by low signal-to-noise ratios, sparse station sampling and complex wave propagation at short periods. We build upon previous work in designing a multiple-event, simultaneous moment tensor inversion scheme for small earthquakes that employs constraints from P-wave polarities, relative amplitudes of P- and S-waves recorded at common stations, and local magnitude estimates. Our method does not require a priori knowledge of a reference moment tensor. High-fidelity polarity and relative amplitude data are recovered using principal component decomposition of clustered-event waveforms. These data are employed within a multi-stage iterative framework to invert for moment tensors and incorporate local magnitude information. Synthetic examples employing as few as four high-quality and spatially-distributed stations yield accurate moment tensor estimates. We demonstrate our approach on a cluster of seismicity near San Juan Island, Washington, USA, within the Cascadia forearc. Our results are consistent with previous characterization of the local stress regime, and support an interpretation of swarm behaviour resulting from migration of fluids originating from dehydration of the subducting Juan de Fuca plate.