Authors

D. Kilb (IGPP/SIO/UCSD)
G.A. Prieto, IGPP/SIO/UCSD
V.G. Martynov, IGPP/SIO/UCSD
F.L. Vernon, IGPP/SIO/UCSD

Project Description
We rank the relative complexity of aftershock sequences using the results from waveform cross-correlation measurements. We purposely use prevalent, yet under utilized, seismic waveforms recorded by networks with non-optimal station coverage (azimuthal gap >225°) to establish their usefulness. We examine five aftershock sequences: two sequences (2812 earthquakes) in Tien Shan, Central Asia, and three sequences (1396 earthquakes) in California. Qualitatively assessing stacks of optimally aligned suites of seismograms from each aftershock sequence we find many stations exhibit coherent signals that are well above the noise. More quantitatively, at each station, and for relatively close earthquakes (<5km horizontal distance; <10km vertical distance), we determine pair-wise waveform cross-correlations of 1-8Hz filtered data windowed 8 seconds around the analyst picked P-wave arrival. For all five sequences, the median percentage of waveform pairs that have cross-correlation values ≥ 0.5 (our assumed threshold for similarity) never exceeds 40%. These relatively low percentages are not a function of catalog completeness levels, aftershock sequence duration, data sample rate or systematically sampling only stations with low signal-to-noise ratios. We can also rule out near station velocity variations as the primary cause of the low correlations because at a single station different sequences yield very different correlation measurements (some high, some low). We suggest the relatively low correlations found in this study (<40% compared with >70% elsewhere) result from either complex earthquake source processes within the sequence (i.e., variable fault orientations or rupture directivity) or an extremely heterogeneous fine scale velocity structure in the source region.