Ellipticity joint inversion vith DC using Dinver 3.3.0. and following relaeases

This forum is dedicated to discuss all problems and suggestions related to the inversion software
Post Reply
luigiV
Posts: 33
Joined: Fri Jul 17, 2020 9:28 am

Ellipticity joint inversion vith DC using Dinver 3.3.0. and following relaeases

Post by luigiV »

Hi Marc,
Using Dinver from Geopsy releases 3.3.0. and following I verified that the ellipticity curve is loaded in degrees (in absolute or signed value) instead of in amplitude as in previous versions. When I check the result, the plot appears with only the theoretical curves and the measured curve is missing. Sometimes it appears but very far from the models and often does not appear at all. I tried to select signed or absolute, but I don't get any positive results. This happens whether you use the complete curve or only the right side (as you suggest in curves without singularities). Can you please tell me why this happens and how I can go about overcoming this problem (provided it is not a bug)? I am attaching a picture for a better understanding.
Thanks a lot
Luigi
Attachments
dinver_ellipticity.jpg
(49.73 KiB) Not downloaded yet

admin
Site Admin
Posts: 537
Joined: Mon Aug 13, 2007 11:48 am
Location: ISTerre
Contact:

Re: Ellipticity joint inversion vith DC using Dinver 3.3.0. and following relaeases

Post by admin »

Hi Luigi,

The first strange thing in your figure is the value of your data curve. Values must be restricted to -90 to 90 deg (or 0 to 90 deg. for absolute value). How did you import the curve into Dinver? 0 deg. corresponds to a null ellipticity, 90 deg. to an infinite ellipticity.

If you have a positive ellipticity and if you select "signed" ellipticity, Dinver will try to fit only with positive ellipticities. With "absolute", the sign is ignored when computing the misfit. Your data curve must be positive but the simulated ellipticities can be either positive or negative.

Best regards,

Marc

luigiV
Posts: 33
Joined: Fri Jul 17, 2020 9:28 am

Re: Ellipticity joint inversion vith DC using Dinver 3.3.0. and following relaeases

Post by luigiV »

Hi Marc,
After your answer I uploaded an ellipticity file calculated by HVTFA as you can see from the attached images. In the first figure the second column of the parser shows the amplitude (atang) of the ellipticity and the corresponding graph. I have associated the ellipticity curves picture of the inversion expressed in the degree scale.
In the second image I modified the input file with the degrees option and the representation is in degrees. The inverted curves were calculated with a parameterization previously tested with Dinver of the snapshot release. As you can see the misfit is absurd. In the third image the input data are always the same but this time the misfit appears acceptable.
The anomalous thing is that the next run, without changing anything at all, the misfit returns exaggerated. In the fourth image I represented the misfit of a sequence of runs from which it is appreciated how the outputs are unreliable and the correct solution is absolutely random and not repeatable with the next run.
Therefore I believe Dinver has a bug that should be fixed.
Do you agree?
Luigi
Attachments
misfit.jpg
(53.66 KiB) Not downloaded yet
dinver2.jpg
(121.45 KiB) Not downloaded yet
dinver1.jpg
(121.11 KiB) Not downloaded yet
dinver.jpg
(124.13 KiB) Not downloaded yet

admin
Site Admin
Posts: 537
Joined: Mon Aug 13, 2007 11:48 am
Location: ISTerre
Contact:

Re: Ellipticity joint inversion vith DC using Dinver 3.3.0. and following relaeases

Post by admin »

Hi Luigi,

Sorry this very very late answer.
I'm a bit lost with the reference to images in your text.
Anyhow, I will comment on each image:

dinver.jpg:
The ellipticity value are all very low for an ellipticity in degrees. Probably in the parsing filter the type of column was not set properly

dinver1.jpg:
The range of values looks better for degrees. What's wrong with the inversion is that dinver tries to fit the curve with a signed ellipticity, even though "absolute" was selected.

dinver2.jpg:
It looks better but unexpected for signed ellipticity.

misfit.jpg:
I have no idea of the target defined for each run. Run 4 and 5 are probably corresponding to target of dinver2.jpg. Other runs to the other targets.

I will check the relative/absolute flag. If you still have the .dinver file, I would like to have a look at it.

Best regards,

Marc

luigiV
Posts: 33
Joined: Fri Jul 17, 2020 9:28 am

Re: Ellipticity joint inversion vith DC using Dinver 3.3.0. and following relaeases

Post by luigiV »

Thanks Marc,
first of all I want to confirm to you that the contradictory results of the misfit.jpg image always refer to the same input model and the strange thing I wanted to introduce was just that. Therefore I thought there was a bug in the program.
I continued the experimentation and found better and congruent answers using only the right side of the peak, but I did not understand why the choice of the parser filter has always been the same. It would be very useful for me to have an exact description of the parser structure to be used with Dinver release 3 and higher.
I also found different misfit significantly using Dinver of release 2.9.0 and Dinver of release 3.2. with the same input parameters. Is there a reason for this or is it my inexperience? Personally, I am more comfortable with version 2.9 or that snapshot that accepts data in amplitude and always output with that magnitude. Is there a practical reason to prefer Dinver from releases 3 ... rather than previous releases?
I am attaching that dinver file of images. Thanks

I would like to take the opportunity to kindly ask you if it is possible with Geopsy to define the azimuth of a source, if it exists, when using FK and therefore understand if the dispersion data will be reliable or if it will be more correct to use the Spac approach.

Thanks for answers
Luigi
Attachments
2812_ell.txt
(3.22 KiB) Downloaded 53 times

Post Reply