Degree
Doctor of Philosophy
Program
Geophysics
Supervisor
Dr. Kristy Tiampo
Abstract
The implications of understanding fault networks as ergodic systems are addressed here by using the Thirumalai-Mountain metric that identifies effective ergodic periods, when long but finite time intervals are considered. This framework was proven to be useful in statistical seismology studies. Initially, it was established that that the metric can be used to quantify seismicity clustering. Periods of effective ergodicity were characterized by the occurrence of declustered seismicity. This interpretation was implemented for synthetic and seismic data from southern California and Canadian mines.
Next, methods used in the estimation of regional seismic hazard were applied to mining induced seismicity. The interpretation of this metric as a measurement of seismicity clustering was fundamental to the adaptation of these techniques to mining seismicity. The latter also provided a better understanding of the mechanism for increased efficiency of hazard assessment methods based on seismic patterns during these ergodic periods.
In addition, this metric was employed in order to develop a robust seismic declustering technique that does not depend on a large number of parameters. Optimization methods and the Gutenberg-Richter law were used as constraints to identify clustered events in a given dataset. This method was applied to a synthetic catalog and seismic data from regions with different tectonic settings: southern California, Taiwan, Switzerland and the Gibraltar arc. Comparison to other declustering methods applied to the datasets did not show the same success due to their parameter dependence.
The effects of location errors on a particular pattern-based seismic hazard assessment technique was also studied. Perturbed catalogs were generated from the southern Californian dataset by adding noise to epicenter locations. Seismicity trends identified by the metric did not change with the increase in noise levels. A combination of the latter and the large number of small events offset the effects of location errors in the performance of the considered method in retrospective forecasts, where no systematic degradation was found. This indicates that these uncertainties do not affect the technique significantly. The same occurred for smaller catalogs.
Finally, remarks on the advantages and limitation of the framework are discussed along with suggestions of future work.
Recommended Citation
Cho, Nelson, "Ergodicity and seismicity clustering with applications in statistical seismology" (2011). Electronic Thesis and Dissertation Repository. 206.
https://ir.lib.uwo.ca/etd/206