Hamiltonian Monte Carlo (HMC) is known to be highly efficient when sampling high-dimensional parameter spaces. This high efficiency can be attributed to Hamilton’s equations, which guide the sampling of the model space. In the case of weakly non-linear problems, this efficiency can be increased even further by linearizing the forward problem. In this study, we exploit this for the purpose of estimating source parameters of a 3.4 magnitude induced event that originated in the Groningen gas field in 2019. In total, we estimate ten earthquake parameters: centroid (three coordinate components), moment tensor (six elements), and origin time. We demonstrate that, in the absence of a sufficiently accurate centroid prior, the linearization of the forward model necessitates multiple initial centroid priors. Here, we consider two sets of initial centroid priors. The first set is based on the known fault geometry in the Groningen reservoir, whereas the second set is obtained by placing initial centroid priors on a uniform horizontal grid at a depth of 3 km (the approximate depth of the gas reservoir). In general, the results from both sets are in good agreement with each other. Most important, however, is their agreement with the geological knowledge of the area: the posterior peaks for model vectors containing a centroid near a major fault and a moment tensor that corresponds to normal faulting along a plane which has a strike almost coinciding with the strike of that major fault.

Timing errors are a notorious problem in seismic data acquisition and processing. We present a technique that allows such time shifts to be detected and corrected in a systematic fashion. The technique relies on virtual-source surface-wave responses retrieved through the crosscorrelation of ambient seismic noise. In particular, it relies on the theoretical time-symmetry of these time-averaged receiver-receiver crosscorrelations. By comparing the arrival time of the surface waves at positive time to the arrival time of the surface waves at negative time for a large a number of receiver-receiver pairs, relative timing errors can be determined in a least-squared sense. The time-symmetry of the receiver-receiver crosscorrelations, however, is contingent on a uniform surface-wave (noise) illumination pattern. In practice, the illumination pattern is often not uniform. We therefore show that weighting different receiver-receiver pairs differently in the inversion allows timing errors to be determined more accurately. The weights are based on the susceptibility of different receiver pairs to illumination-related travel-time errors. The proposed methodology is validated using both synthetic data and field data. The field data consists of recordings of ambient seismic noise by an array of stations centered around the tip of the Reykjanes peninsula, southwest Iceland (some of these stations exhibit time shifts of an unknown nature).

We report on the extraction of deep ocean travel time variations from time-lapse cross-correlations between a hydrophone station and a three-component broadband seismometer. The signals we cross-correlate in this study result from repeated activity by the Monowai seamount, one of the most active submarine volcanoes of the Tonga-Kermadec ridge. In particular, we introduce a specific workflow to exploit repetitive hydroacoustic underwater source activity, which we detail to such extent that it serves as an example (or “cookbook”). For this reason, we have made the source code publicly available. The workflow proposed in this study (i) overcomes differences in instrument sensitivity and sample rates, (ii) involves the selection of eligible cross-correlations based on a source activity criterium as well as slowness analysis, and (iii) extracts the travel time variations in distinct frequency bands. In our case, the two frequency bands are 3-6 Hz and 6-12 Hz. We find that the estimated travel time variations in both frequency bands consist of a complex periodic pattern superimposed on a distinct long-term linear trend. This long-term linear trend is decreasing, which we attribute to increasing water temperatures along the propagation path of the hydroacoustic signals.