Few questions for FK analysis

[Uki]

Hi Marc,

I have few questions for the parameters and output data while performing FK analysis:
1. Is it possible to obtain the value of F-K spectra (the beampower value in time window browser)?
2. What is the 'valid' inside the output max file stands for?
3. What does the 'weight' mean in gphistogram after added curve?
4. In some cases, there is gap (frequencies with very few blocks) could be seen on the dispersion curve. For example, around 1.24 Hz in the attached figure. May I ask for your opinion?


Best regards,
Uki

[admin]

HI Uki,

1. Is it possible to obtain the value of F-K spectra (the beampower value in time window browser)?

No apparently there is no entry in the context menu to export the values. The code should be slightly modified

2. What is the 'valid' inside the output max file stands for?

It is a flag set to 0 or 1. Lines with 0 are ignored in the computation of the histogram but that are not removed for future exports.

3. What does the 'weight' mean in gphistogram after added curve?

It is set to the number of samples used to compute the statistics. It can be further used to merge various curves obtained from distinct jobs with varying number of windows.

4. In some cases, there is gap (frequencies with very few blocks) could be seen on the dispersion curve. For example, around 1.24 Hz in the attached figure. May I ask for your opinion?

Did you try to browse the FK spectra at that frequency? Peaks outside MIN_V and MAX_V? The probabily density plots shown in your figure, are they computed at 1.24Hz? I have the impression that it is at a higher frequency because the first peak is below 1000 m/s. Does 1.24 Hz is where there is only a single bin around 1200 m/s? Maybe all values at that frequency are very close. Definitively, having a look at the FK spectra at 1.24 Hz might be instructive.

[Uki]

Dear Marc,

Thank you for your prompt reply.

For the attached dispersion curve figure, the probability density is not computed at 1.24 Hz as you said. Sorry for the confusion.

I have checked the FK spectra, and the reason of the gap around 1.24 Hz is just as you said, the maximum peaks of FK spectra have the velocities above MAX_V, so they might be filtered and not be included into the output max file.

Also, I find something interesting about the peak on FK spectra.
Here I attach two figures, they are the FK spectra of a dataset utilizing single component HRFK. The parameters for calculation are the same except one allows 1 peak on FK spectrum (FK_max1peak_relative0%.png), and another one allows two peaks (FK_max2peak2_relative0%.png). No threshold (absolute or relative) is utilized. The yellow block shows the information relates to the peak.
As you can see the peak positions (red X) between 2 figures are different.
In FK_max2peaks_relative0%.png, for some reason, the peak with 2nd strongest power been chosen, and no peak been picked at the position with the strongest power. Moreover, although the maximum number of peaks is set to 2, only 1 block appears two peaks within the 28 blocks (fig3.png). I took a quick look and I believe this happens when the power of outside peak (larger wavenumber) is larger than the inside peak.

Best regards,
Uki

[admin]

Hi Uki,

I think that the strange movement from one peak to the other is linked to a bug in the display that I fixed at the beginning of July (unsure, vacations since then). If you are running under Linux, I would be curious to see the difference if you install the latest development through git repository (branch 3.6). It includes also the choice of an experimental derivative based algorithm to search the peaks (option in FK parameters).

By the way, did you check the peaks in the max file at that frequency? Is there one or two peaks?

Read you,

Marc

[admin]

Is it possible to obtain the value of F-K spectra (the beampower value in time window browser)?

No apparently there is no entry in the context menu to export the values. The code should be slightly modified

Now, all grid plots have an export value option in the context menu.

[Uki]

Hi Marc,

I am running Geopsy (3.5.2) under windows 10 (22H2).

There are two velocities (two rows) for the two peaks on FK spectrum in the max file.
After few more testing, I believe the reason that there was only one peak on the FK spectrum in FK_max2peaks_relative0%.png is because the vmax in FK gridding.
Here is what I am thinking, please correct me if I am wrong: When N peaks are allowed, geopsy first picks N maxima on the FK spectra and then filter the result within the v range (vmin to vmax). So the peaks without the v range will be excluded and not plotted on the FK spectrum, which makes the peak on the figure looks like shifting. However, when only 1 maximum is allowed, the peak with maximum power will be picked and plotted on the FK spectrum, no matter whether it's within or without the v range. In this case, if the calculated velocity is without assigned v range, it won't be written into the max file.

I couldn't find the option in context menu to export value. Do I have to first export the result to a page file?
(viewtopic.php?p=299#p299)

Best regards,
Uki

[admin]

min and max velocities are used to define the grid search (translated into max and min wavenumbers). N maxima are effectively picked using the grid search. Upon saving the results, the min and max velocity are again checked. I'm unsure whether this second check is really useful. Close to the wavenumber limits the second condition might be useful. In your case, all peaks are well inside the wavenumber limits and such case should not happen. 3.5.2 does not contain the required corrections for showing all peaks. You have to compile the binaries from a source tree downloaded with git, to be on the latest branch where I did all recent modifications. I recently wrote a tutorial to compile under Windows but it is easier under Linux.

The export of grid values in a simple way (not the tricky one detailed in the reference post) is only available from branch 3.6.

Best regard,

Marc