Studies on earthquake location and source determination using coda waves

Abstract

The majority of existing seismological techniques ignore high frequency coda that result from scattering and focus on early onset body waves to model earthquake location and source properties. A recent technique known as coda wave interferometry utilises the cross correlation of coda waves to constrain changes in source properties between earthquake pairs. This thesis focuses on extending coda wave interferometry by designing mathematical frameworks and practical algorithms which enable the use of coda in studying the properties of earthquake clusters. Synthetic experiments are designed to quantify the constraints provided by coda wave interferometry and two corrections introduced which extend the range of applicability from separations of 300 to 450m for 1 to 5 Hz filtered waveforms. It is shown that wavelength normalisation of separation leads toa uniform behavior across frequency bands and that coda wave constraints are inherently probabilistic in nature. A Bayesian framework is introduced to determine the probability for actual separation given coda wave estimates and a technique proposed for combining coda wave and travel time data. It is shown how a posterior density function describing the joint relative location of an earthquake cluster can be built using the pairwise Bayesian framework. This joint posterior is studied using direct search, ensemble inference and optimisation under a range of conditions for earthquakes in Western Australia and California. Combining coda and travel times leads to the best solution in most cases. It is shown however, that coda waves significantly enhance location constraints when the uncertainty associated with travel times exceeds half the event separation. Furthermore, coda waves are demonstrated to provide valuable information in poor recording situations with few stations and can be used with as little as one station. In contrast, travel time techniques require multiple stations and good azimuthal coverage. This feature of coda waves to succeed with limited data is its greatest strength, lending itself to use in intraplate regions where station density is sparse. An extension of coda wave interferometry is developed which relates the change in orientation of two identically located double couple sources to the correlation of their coda waves. Synthetic studies demonstrate that this relationship is accurate for a range of source mechanism changes and that unlike the separation theory, does not depend on frequency. This suggests that it is feasible to differentiate influences of source displacement and orientation directly from the coda and opens the possibility of a combined coda wave interferometry for source separation and mechanism variation.