Toward fully automatic earthquake detection and processing for tomography in the Hengill area
- Plats: Hambergsalen, Villavägen 16, Uppsala
- Doktorand: Wagner, Frederic
- Om avhandlingen
- Arrangör: Geofysik
- Kontaktperson: Wagner, Frederic
This thesis focuses on the automation of seismic data analysis, in particular, event detection, quality assessment of detected events, and preparation of an earthquake catalogue for seismic tomography.
The developed event detector uses back-propagation and stacking of a seismic trace attribute with a known velocity model to detect and locate events. A four-dimensional volume in space and time is probed for local maxima of coherently stacking signals. These local maxima define event location and origin time. Application of the detection algorithm to data from a dense 26-station 3-component seismic network in the Hengill area, SW Iceland, produced an increased true-to-false detection ratio compared to the local detection routine.
The detected events were analysed using inter-event cross-correlation with a manually picked reference catalogue to determine their similarity with real events. Automatic P- and S-phase picks were derived using the time delay information from highly correlated events. Relocation with the determined phase picks improves hypocentre uncertainty. A multi-stage selection process is implemented to categorise the detected events into different classes of varying priority for a potential manual analysis. Depending on the used parameters, the top quality category of events can be used in e.g. local-earthquake tomography without manual inspection. Iterative application of the algorithm improved the reference catalogue by almost 40% with events of at least equal quality.
The final local-earthquake tomography with the updated reference catalogue confirms the success of the implemented workflow. The resulting Vp, Vs, and Vp/Vs models show structures that can be associated with the local geothermal activity. A higher resolution and extended ray coverage was achieved compared to previous tomographic studies. Double-difference location of the events using differential times from waveform correlation significantly improved event hypocentres revealing detailed fault geometry in the known seismicity pattern. A preliminary double-difference tomography shows promising results for high resolution imaging.