Current status of SPAC functionality

[iandkelly]

Hi Marc,

I've been scouring through the MSPAC documentation on GeopsyWiki (https://www.geopsy.org/wiki/index.php/MSPAC) and through previous forum posts to try to piece together the current functionality/limitations of SPAC processing in Geopsy, but several aspects remain unclear to me. Would you kindly clarify from the below questions:

1. I understand that 3C SPAC data can be processed by spac2disp via the -VRT options, but the output targets (-disp.target and -forward.target) aren't displayed by the spac2disp GUI, and don't seem to display in Dinver either. Is it presently possible to visualise the dispersion results from 3C SPAC, or am I only able to generate the 3C autocorrelation curves?

2. Following from above, is it possible to invert the dispersion results from 3C SPAC in Geopsy? Or alternatively, is there an avenue to directly invert the autocorrelation curves in Geopsy? My current testing seems that you end up with the .target, -disp.target and -forward.target files, but then can't do anything with them afterwards or make any subsurface interpretation.

3. I recall it mentioned in the forums that the kmin and kmax parameters for dispersion curve picking of 1C SPAC data in spac2disp are arbitrarily defined compared to considerations of array response functions for FK processing - am I correct in this assumption? Or are there certain resolutions I should consider?

Thanks for all your work into Geopsy and looking forward to hearing back from you soon!

Best,

Ian Kelly

[admin]

Hi Ian,

To formulate an answer, I checked the code (version 3.5.0-preview) with a synthetic case (model M2.1 from SESAME project). It should work the same with 3.4.2.

1. As you noticed,

Code: Select all

spac2disp -VRT array2-3C.target

produces two .target files: -disp.target and -forward.target. The first one is the dispersion curves for Rayleigh and Love waves (see attachment forward-target.page). The second has 6 curves per ring: 3 are the data curves, the same as in the input target file (array2-3C.target), the last 3 do not have error bars. They are the vertical, radial and transverse autocorrelation curves simulated for the best fits (see attachment forward-target.page).

In forward-target.page, you can see the Rayleigh and Love curves in red and blue, respectively. The dotted lines are the true dispersion curves. The modes are correctly retrieved except around 4 Hz for Rayleigh where the first higher mode disturbs SPAC processing. Processing with RTBF of the same case shows that higher modes are present at this frequency. SPAC is not able to distinguish and separate modes.

In forward-target.page, the simulated curves are in red. The data curves are in black. At low frequency a good match is observed. At higher frequency, higher modes and aliasing make the situation more chaotic.

spac2disp provides also at the end of the process the alpha curve: variation of the proportion of Love and Rayleigh waves as a function of frequency. There are also plenty of .report produced. They are the results of the inversion at each frequency. For example, at 9.9 Hz, you can plot the ensemble of models generated during the inversion

Code: Select all

gpdcreport frequency_9.93847.report -pm | figue -d -x-column 2 -y-columns 3 -yerr-columns 5

The result is shown in frequency_9.93847.page

2. You can invert -disp.target as regular dispersion curves. The two polarizations can be included in the inversion. Note that a manual cleaning of the dispersion curves is certainly required. The direct inversion of the 3C autocorrelation curves is not implemented. If Love and Rayleigh phase velocities are available with the usual dispersion curve inversion, the inversion of 3C curves would require additional parameters for alpha, which shape is relatively random. If there are 100 frequency samples, 100 additional parameters would not be feasible with the neighbourhood algorithm.

[luigiV]

Hi,
first of all thank you for giving me the opportunity to implement the little I know about the Spac 3C. So far I was stuck on the error message that appears when using spac2disp from the Geopsyland link.
I tried to process with 3.4.2 a file of the tutorial database (lep_ring01_3C.gpy) with the possibility therefore to compare the result with the 1C one published in the Mspac documentation. I used the same inputs reported in the tutorial and the final result is reproduced in lep_ring.page.
There are no Love waves as demonstrated by FK_radial_Capon: the blue curve is therefore unreliable, while the one referred to Rayleigh is confirmed in the sampled area on the frequency-slowness PDF plot in terms of minimum (6Hz) and maximum (20Hz) frequency and also by FK_vertical_Capon output.
Marc,your help would be very useful for a correct reading of the alpha plots and then how to modify the gpdcreport script to get the velocities instead of the slowness.
Thanks

Luigi

p.s. In FK.page I have compared the results obtained with RTBF and Capon processed with the default parameters for the different types of process. Beyond the fact that there are no Love dispersions, it seems to me that the quality of vertical Capon is much better than the RTBF one. Is it possible to comment on this?

[admin]

Hi Luigi,

In lep_ring.page, there are two missing external images:
C:/lavori/lep_ring/disp.jpg
C:/lavori/lep_ring/disp+fk.jpg

Can you edit your message to add them?

[admin]

To get velocities instead of slowness the preferred method is to change the scale type in the plots. Switch to "Inverse log". Because internally slowness is used (somehow an arbitrary and historical choice), it is better to keep slowness. Comparison with other results will be easier. If you prefer velocity for plotting (which I do also), simply change the axis scales.

[luigiV]

Hi Marc,
sorry for the delay, but I did not notice your reply.
I don't know if I understand your request (I see the page file complete with all the plots) and therefore I am attaching the jpg where all the plots are included.
I don't understand how this can happen. Maybe the jpg files cannot be saved in a sent page file?

I also took the opportunity to replace the radial Capon plot with the transverse Capon plot.

Regards