Last year I published my script on VSP modeling and processing. And now when I’ve looked back at the script, I’ve found so many errors. I’ve updated the script and added one or two features, if anyone of you have free time, maybe you can help me to improve this scripts.
First major issue, that actually I had realized it last year was about the synthetic modeling using Ela2D. I don’t know how to set it up properly for VSP geometry besides setting the receiver to be increasing with depth. Because, the synthetic model produced the waveforms too slow. Maybe “too slow” is not the right term, so basically, the issue was when I processed the synthetic datasets for flat subsurface layering, the upgoing does not come out as a flat reflection. So maybe the direct P modeling/computation that I set in the script was not correct. I haven’t solved this issue, but I probably want to switch to SOFI2D from Univ. Karlsruhe for Synthetic Modeling.
Second major issue, that I haven’t solved is scripting for waveshaping deconvolution. Which is VSP style of up by down deconvolution. Seismic Unix provided sushape in doing Wiener-Levinson waveshaping deconvolution, but somehow I failed to scripted for VSP deconvolution. CUrrently I am using supef, prediction error filtering, which I think sufficient enough in my case.
Last, though I think overall scripting is sufficient for the exercise of processing zero offset VSP, I still feel uncomfortable with the quality of the results. Things such as, using trace numbering for index in median separation looks not robust, the Seismic Unix band pass filtering looks changing the shape of wavelet around break time too much. So maybe if I have time, I can improve it again.
So, if you have time, have a look at my note on VSP processing and modeling using seismic unix here www.gm2001.net/latihanvspsu and www.github.com/fahdi104/latihanvspsu. I’ve also written some description in this page https://inversi.wordpress.com/vsp-modeling-processing-with-seismic-unix/