I think I should write these ideas, while these things still running around inside my head. Though, again this idea might seems impossible to be done, hehehe
1. Anisotropic based on Shear Wave Splitting analysis on Zero Offset VSP.
There were several attempts on this one. Internally and also externally outside my current company. What I have red so far about fast shear and slow shear extraction using parametric wavefield decompositions (Parametric wavefield decomposition and applications, Leaney et al, 1990 ).
But today, I read a recent paper from Curtin Univ group, they attempted to run semblance-wise analysis to extract shear wave splitting. My understanding as of now regarding this paper are,
a. You project your horizontal component to several azimuth,
b. Using the projected horizontal component, you run semblance analysis for multicomponent linear traveltime moveout velocity analysis. The main idea is to estimate the dominant apparent velocity for a given polarization direction by measuring the coherency of the seismic signal of a large number of events as a function of the apparent velocity. So, this look like semblance analysis on the presence of a large number of shear wave events with the same apparent velocity.
Usually and ideally, we will need multi offset walkaround VSP in order to properly measured anisotropic parametr based on shear splitting phenomenon. But this paper reported that their attempt can be done by analyzing many low-amplitude PS mode converted waves on zero offset VSP.
To me this is doable, since projection and semblance is available within our current software. But lets see whether this is really doable or not.
Detail of the methods can be seen in this paper, Estimation of azimuthal anisotropy from VSP data using multicomponent S-wave velocity analysis, 2011
2. Full waveforms inversion of VSP data
This is a very optimistic idea🙂 As of now, the full waveforms inversion is regarded as one of the best solution to image complex structures. But the idea of inverting the whole waveforms of VSP data in this project is just to approximate higher velocity field information. Because normally, what we invert from VSP data is just the traveltime, which yields lower frequency information.
Why this is doable🙂 a lot of forward modeling code is available for research purposes (SeismicUnix, Crewes, SOFI, etc). Lets forget about the sophisticated method of iterative solver, frequency domain modeling etc etc. Lets just use what is available for the forward modeling, and focus on the inversion.
Lets try it with local optimization approach, and then maybe if I suddenly I struck by lightning and become a genius, we can try to run the global optimization approach for the inversion.
The guide paper for this is Nonlinear Inversion of Seismic Waveforms, by Marwan Charara, Christophe Barnes and Albert Tarantola.
Anybody care to share their thoughts on my project?