Questions about optimal strategies for Seismic Inversion

[kerim371]

Hi,

I'm starting using SPECFEM for FWI purposes and I consider using seisflows for that.

1. In seismic exploration when we calculate wavefields we usually don't want to calculate it on the whole field but rather we use some part of the mesh restricted by the survey i.e. some area within source and receiver positions plus some margin.
   How this can be done in seisflows? Because if we have for example survey located on xmin=1km and xmax=3km and the whole mesh is 100km, it is not optimal to calculate wavefileds for the whole mesh, there is no receivers.

2. Another question is that when to calculate FWI we need to somehow store forward wavefileds to do some mathematics with adjoint wavefields. As far as I understand SPECFEM stores forward wavefields on the disk. Is there a possibility to keep them in RAM? Because storing them on disk may require too much time.

Thank you in advance!

[bch0w]

Hi @kerim371, I think these are questions more suited to SPECFEM than SeisFlows, and I would suggest posting in the relevant SPECFEM discussion or issue tab. Some short answers to your questions:

1. You could shorten your simulation so your wavefield does not make it to the outer regions? Or optimize your mesh to fit your survey domain
2. This would probably require modifying the SPECFEM source code to do what you need with the forward wavefield in RAM. This question is better suited for the SPECFEM groups, however.

[kerim371]

Thank you! Sufficient for me for the now

[kerim371]

@bch0w hi,

I've done some first experiments with seisflows, I've watched a some videos and new questions had arised.

As I'm from seismic exploration I'm interested in tasks when there is a some survey geometry wich consists of a number of sources and receivers "moving" along 2D line on the surface.

With specfem I can set positions position of a singe source with SOURCE file and positions for the receivers in a STATIONS file.
For both of them I have a single Par_file where I can set up the meshing and modeling parameters for that particular source.

If I have many sources then from my understanding I need to have many Par_file and SOURCE/STATIONS files. So for each source I need to have DATA and DATA_OUTPUT folders.

Does seisflows allow me to somehow automate the process of creating such folders with mentionned files within for each source or do I have to do that manually i.e. for each source create DATA, DATA_OUTPUT dirs and properly setup Par_file and SOURCE? or there is some other ways of doing computations for different source position?