[MSNoise] 1Bit and Whitening: Order of operations

[Francesco Civilini]

Hello all,


I would like to suggest an option in MSNoise to change the order of operations from Bandpass -> 1Bit Norm -> Whitening to Bandpass -> Whitening -> 1 Bit Norm in the s03_compute_cc.py routine.

I have observed that for datasets where there is significant correlation in the higher frequencies by a cultural source (for example machinery near two stations), but the frequencies of interest are lower, the resulting correlation will be completely dominated by the high frequency signal if 1Bit Normalization occurs prior to whitening.


This could lead inexperienced users with noisy datasets to think that there are no low frequency signals propagating across their array. By doing 1Bit normalization prior to Whitening, it will no longer be possible to filter the cross correlations to further narrow the target frequencies, which is not the case if Whitening occurs before 1 Bit.


A workaround for the current version of MSNoise is to change the original bandpass to a frequency range that excludes the problem frequencies, but this requires prior knowledge of what the cross correlations will look like for a wide frequency band.


Does anyone have any personal examples for datasets that support or contrast these observations?


Thanks,

-Francesco


Francesco Civilini, M.S.
Geophysics PhD Candidate
Victoria University of Wellington
Cotton Building Room 505
Wellington 6012, New Zealand
(027) 868-5939
Francesco.Civilini at vuw.ac.nz<mailto:Francesco.Civilini at vuw.ac.nz>
www.francescocivilini.com<mailto:Francesco.Civilini at vuw.ac.nz>