Use of coda wave interferometry for estimating properties of earthquake sources
David Robinson1, Malcolm Sambridge1 and Roel Snieder2
1 Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia
2 Center for Wave Phenomena and Department of Geophysics, Colorado School of Mines, Golden CO 80401, USA
The waves arriving later in a seismogram arise from scattering and are known as coda waves. Most techniques for studying earthquake source properties such as location and mechanism do not use the coda. Coda wave interferometry relates the variation between two earthquake sources and the cross correlation of their coda waves. In this project we are researching the applicability of coda wave interferometry (CWI) for constraining the location and mechanism of double couple events.
Snieder (2005) demonstrate how the separation between two double couple events with identical source mechanisms can be estimated using CWI. We have conducted numerical experiments in stochastic heterogeneous media to explore the range of applicability of CWI for estimating source separation. We observe that CWI estimates of separation are within one standard deviation of actual separation when the perturbation is less than one fifth the dominant wavelength. Moreover, we have demonstrated how CWI estimates of separation should be interpreted with the aid of a probability density function (PDF) which computes the probability of different actual separations for a given CWI estimate (Robinson, 2007a). We have also extended existing CWI theory to show how a change in the source mechanism between two identically located double couple sources can be estimated from the correlation of their coda.
We are currently comparing the performance of CWI for contsraining earthquake locations by using data recorded in the Paradox Valley, Colorado. The earthquakes that we are studying were induced as part of local salinity control which includes injection of saline water to depths exceeding 4km. These induced earthquakes were recorded in an ideal recording situation with good azimuthal coverage. Therefor, existing location techniques are expected to perform well. We can use these data to explore the usefulness of CWI in locating earthquakes. In particular, we can resemble poorer recording situations by randomly discarding data and we can compare the pereformance of CWI against traditional location techniques which are known to perform badly when azimuthal coverage is poor. The figure demonstrates that CWI has the potential to produce accurate estimates of earthquake separation for each station separately and suggests that the technique may be useful when azimuthal covereage is poor.
Figure 1. CWI estimates of separation as a function of sliding time window (white lines) for a pair of events from Paradox Valley which previous modelling has suggested are 100 m from on another. Each white line represents data from a different channel-station combination. The yellow error bar represents the mean and +/- one standard deviation when data from all chanell-station combinations are included.
Snieder, R., and M. Vrijlandt, Constraining Relative Source Locations with Coda Wave Interferometry: Theory and Application to Earthquake Doublets in the Hayward Fault, California, J. Geophys. Res., 110, B04301, 10.1029/2004JB003317, 2005.