search

bgoutorbe / seismic-noise-tomography

Python framework for seismic noise tomography
169 stars 89 forks source link

Test cases #13

Closed ThomasLecocq closed 9 years ago

ThomasLecocq commented 9 years ago

Hi there !

I'm trying to run some FTAN on my own cross-correlation functions (CCF) computed using MSNoise... so far, I think I have managed to mimic it, but I'm not quite sure of the result.

Could you provide some test-case CCF files and their FTAN (image, vgarray) ?

Cheers,

Thomas

bgoutorbe commented 9 years ago

Hi Thomas,

sure, I'll extract some pairs of stations on which I'm working, just give me a few days.

Cheers Bruno

On 06-11-2014 12:25, Thomas Lecocq wrote:

Hi there !

I'm trying to run some FTAN on my own cross-correlation functions (CCF) computed using MSNoise... so far, I /think/ I have managed to mimic it, but I'm not quite sure of the result.

Could you provide some test-case CCF files and their FTAN (image, vgarray) ?

Cheers,

Thomas

— Reply to this email directly or view it on GitHub https://github.com/bgoutorbe/seismic-noise-tomography/issues/13.

bgoutorbe commented 9 years ago

Hi Thomas,

I've just added a file 'test case.zip' in the repository, which contains a CCF (vs time), the raw/clean FTAN images, and the clean dispersion curve (group velocity vs period).

Please note that the code is still experimental. The FTAN is OK, but I've had a hard time defining formal criteria to extract dispersion curves, and I'm still not quite satisfied with the algorithm. I seek the dispersion curve that minimizes (function pscrosscorr._vg_minimizes):

minus sum(amplitude) + sum(dvg^2),

with dvg the velocity difference between two periods, for the curve to ride through maximum amplitudes while minimizing discontinuities. But of course it strongly depends on the amplitude level, so that in regions of low amplitudes (which tends to happen towards larger periods) the smoothing term has too much strength. I'm thinking of using an amplitude normalized to unity at each period, and an additional parameter to control the relative strength of the smoothing term which will have to be fine-tuned.

Cheers Bruno

ThomasLecocq commented 9 years ago

Excellent! I'll have a look at this super rapidly ! Cheers!