The Utility of Signal-to-Noise Ratios to Filter GNSS Instantaneous Velocities for Rapid Earthquake Characterization

Phase information recorded by GNSS receivers in the near-field of large earthquakes can be used to determine instantaneous GNSS velocities (instavel) for earthquakes or for rapid deformation processes. This technique is different from traditional GNSS positioning in that we use subsequent phase information to directly solve for receiver velocity. The resulting products can be integrated into seismic static-offset estimation, moment tensor inversions, or peak-ground velocity (PGV) computation to assess shaking intensity. However, for effective characterization of earthquakes, it is necessary to either use GNSS stations that are minimally affected by cycle slips, clock biases, and noise levels that are comparable to the event signal itself, or are corrected for the same. Empirical relationships used to, for instance, infer earthquake magnitude from PGVs generally do not assess the noise levels in the velocity time-series. This can systematically bias PGVs and derived products. This study investigates the utility of signal-to-noise ratio (SNR) as a tool to select high quality sites for PGV estimation for any given earthquake. We use data from several recent large earthquakes, including the 2019 Mw 6.4 and Mw 7.1 Ridgecrest and the 2021 Mw 8.2 Chignik earthquakes, to elucidate the factors affecting SNRs and the importance of filtering GNSS stations based on these values to estimate PGV-based moment magnitudes. The SNRs are computed for both the event signal and the peak value of the signal with respect to the background noise. The SNRs show a vivid correlation to hypocentral distances, site-effects, and possible basin-responses. For the 2021 Chignik event, these values for corresponding GNSS stations also agree with the general limits of the earthquake rupture. Presently, the method is partially automated and can be used for rapid earthquake characterization, and the resultant PGVs could be integrated into automated ShakeMap calculations.