High resolution earthquake location using full-waveform analysis

Abstract

The location of earthquakes in the subsurface (so-called hypocentres) provides detailed information on the fault(s) that ruptured. This information, in turn, is crucial in development of our understanding of the structure – and risks – associated with the fault system. In this project, we developed, tested, and interpreted earthquake relocation techniques on three data sets. The first data set is from a CO2 sequestration site called Aneth Field in Utah (USA). The second is recorded on a seismic network in the vicinity of the Alpine Fault. The final data set used consists of the aftershocks from the Canterbury earthquakes sequence from 2010/2011.

In recent times, a method for earthquake relocation based on the difference between travel times from earthquakes called “double-difference technique” has become popular in the exploration geophysics and the earthquake seismology community. In our work, we precede the relocation of closely related events by an analysis called hierarchical agglomerate clustering, a general method to group based on a user-defined criterion. Previous work in our group used the criterion of the power spectrum similarity of earthquakes to define sub-clusters of events.

The first result in this project was obtained by Prabhat Shrestha as part of his graduate studies. He was able to reproduce the sub-clusters observed in the Aneth Field comparing the full waveform data in the time domain, as opposed to the power spectrum. Furthermore, he used the confirmed sub-clusters in double difference relocations to refine the fault information obtained.

Secondly, Dr. Carolin Boese led efforts in the central Alpine Fault and Canterbury regions to examine relocations of hypocentres. Results based on clustering and double difference relocations revealed previously unknown structures in the vicinity of the Alpine Fault, but the interpretation of relocations in Canterbury events is part of ongoing work. 

(Ref: "Cross-correlation-based detection and characterisation of microseismicity adjacent to the locked, late-interseismic Alpine Fault, South Westland, New Zealand" published by Earth and Planetary Science Letters 457 (2017) 63-72 - Calum J Chamberlain, Carolin M Boese, John Townend. - http://dx.doi.org/10.1016/j.epsl.2016.09.061)

---

Technical Abstract

Careful analysis of earthquake hypocentres can provide detailed insight in the faults that are involved. In this project we analysed ways to extract additional information about the subsurface structure from clustering hypocentres using hierarchical clustering combined with double difference relocations. Previous clustering results were obtained from variations in the power spectra of events, but we now confirmed that equivalent sub-clustering can be done directly from an analysis of the full waveform time-series. These results were reported in the MSc thesis of Prabhat Shrestha, and a preprint of a submission to the journal Geophysics.

Next, high-accuracy earthquake locations were obtained using two datasets from dense seismic station networks: one set of data was recorded prior to and during drilling of a ~900 m deep borehole aimed at the central Alpine Fault in the Whataroa River Valley, while the other dataset represents the aftershock sequence of earthquakes near Christchurch following the M7.1 Darfield earthquake in 2010.

For each region the data were processed, cross-correlations of earthquake pairs were calculated using the full waveforms, and different relocation strategies were applied. For the central Alpine Fault region a set of so-called template-earthquakes were selected. They were relocated through solving the double-difference relocation equations by using singular-value-decomposition. These templates were then used to detect previously missed micro-earthquakes using matched-filter detection. The templates and detected events were relocated together to investigate the upper crustal deformation in the vicinity of the Alpine Fault. This indicated two distinct groups of events which were further investigated in terms of their faulting characteristics (by analysing focal mechanisms). The results have been compiled for a publication entitled “High precision cross-correlation-derived detection and relocation of background microseismicity adjacent to New Zealand's Alpine Fault” and will be submitted to the journal Earth and Planetary Science Letters.

A subset of 1250 aftershocks was selected at the eastern termination of the Greendale fault that ruptured in the M7.1 Darfield earthquake to investigate the fault's interaction with adjacent structures. This zone lies between the Darfield fault and the fault that ruptured in the M6.3 Christchurch earthquake ca. 20 km to the East five months later. Different clustering techniques (hierarchical agglomerative clustering based on distance, cross-correlation coefficient, time and magnitude) were applied to identify structures in the aftershock distribution. These also provided objective support for clustering resolved by double-difference relocations. While some smaller features could be seen from both approaches, no significant larger structures could be resolved to date. This may be a result of chaotic stress release in this zone of stress concentrations as the result of the rupture of the Greendale fault. A principal component analysis has been undertaken but did not reveal any new features either. Ongoing research is now focused on the faulting information obtained from focal mechanism solutions of this zone. This may reveal how deformation on adjacent clusters is related.