Velocity
Model From modeling result of using 15Hz zero-phase ricker wavelet, we found inconsistency in time between synthesized seismogram and real seismic. Overall amplitude match is encouraging. We speculated that the reason cause time off is because constant density used in deriving velocity from inverted impedence model. Also time to depth conversion for the model is using velocity from one sonic log velocity. These processes will definitely introduce inconsistency. In order to test if the plane wave simulation can image geological structure correctly and satisfactorily, Wei worked to improve his inversion by incorporating neighborhood traces (9 traces used) to derive velocity model directly rather from the inverted impedence. Then using inverted velocity to do time to depth conversion to assure consistency. This page will analyze modeling result from newly computed velocity model.
Velocity
Model
Here we show 3 sections for inline, cross line and time speperately. Computed velocity model is in 65x41x20ft space. Since forward modeling requires model is in 20x20x20 space, thus 4D module volume interpolater was used to interpolated the computed model to required space.
Figure 1. Inline section of original and interpolated velocity model
Figure 2. Inline section of original and interpolated velocity model
Figure 3. Depth section of original and interpolated velocity model
2 Interpolated velocity cubes will be input to the FEM modeling program of Dr. Teng.
Wavelet
Wavelet will usually 15Hz zero-phase ricker wavelet used in previous modeling experiment. Although dominant frequenyc of 1994 seismic is lower than 15 Hz, but bandwidth of real seismic and seismogram is very close in both 1988 and 1994 cases.
Modeling
result
Computed seismograms are normalized against the corresponding real seismic amplitude in amplitude and frequency. Frequency process is basically applying bandpass to the real and synthesized seismic trace by trace, the bandwidth for 1988 and 1994 survey are same, [0, 35]Hz. The amplitude process is simply a scaling factor applied to the synthesized seismogram since input wavelet is not calibrated against the real seismic.
Figure 4. Frequency spectra and amplitude
histograms of 1988 seismic and seismogram
Figure 5. Frequency spectra and amplitude
histograms of 1994 seismic and seismogram
Figure 4 and 5 are power spectrum of real seismic and synthesized seismogram of 1988 and 1994 seismic surveys. From figure 4, we can see real seismic and seismogram have almost identical power spectrum, but it is not the case of 1994 survey, principle frequency of real seismic is different from the synthesized seismogram. In both cases, 15Hz zero-phase ricker wavelet is used for modeling simulation.
Comparison
between seismic and modeled one
First we pick some inlines, cross lines and time slices from real seismic and synthesized seismogram and put them side by side. In inline and cross line cases, seismic and modeled seismic are mirrored to show how close the time is.
Figure 6. Line 1775 of real seismic and modeled
Figure 7. Line 1819 of real seismic and modeled
Figure 8. Line 1850 of real seismic and modeled
Figure 9. Cross Line 3845 of real seismic and modeled
Figure 10. Cross Line 3884 of real seismic and modeled
Figure 11. Cross Line 3937 of real seismic and modeled
Figure 12. Time slice at 3028ms of real seismic and modeled
Figure 13. Time slice at 3200ms of real seismic and modeled
Figure 14. Time slice at 3404ms of real seismic and modeled
By looking at inlines, cross lines and time slices shown above, we can see very close time-match, genearlly modelled seismic shows little lower seismic event angles comparing with real seismic data. That is what should be expected in plane-wave simulation. Time migration will reposition seismic events upward, this will increase dipping angles.
Wiggle
Comparison between seismic and modeled one
Also we plot some of inlines, cross lines and time slice in overlayed wiggle plots since wiggle plot can show alignment between peaks and troughs between real seismic section and modelled seismic section. Time difference will be clearly seen from the wiggle plot. RED ones are modelled seismic, and BLUE ones are seismic. Only 1988 seismic and modelled seismic shown here.
Figure 15. Wiggle plot of line 1775
of real seismic and modeled
Figure 16. Wiggle plot of line 1850
of real seismic and modeled
Figure 17. Wiggle plot of cross line
3845 of real seismic and modeled
Figure 18. Wiggle plot of cross line
3937 of real seismic and modeled
Figure 19. Wiggle plot of time slice
at 3028ms of real seismic and modeled
Figure 20. Wiggle plot of time slice
at 3200ms of real seismic and modeled
From wiggle plot, we can see amplitude and time is really close, but there still exists differences, they will be seen in difference images shown below too. Migration probably will fix part of time difference. But amplitude will only be fixed through rock properties.
Comparison
by differencing
To furthur compare seismic and modelled seismic, we difference seismic and modelled seismic directly. Figure 21, 22, and 23 shows one of inline, cross line and time slice. There exists fairly significant difference, this can be easily seen from wiggle plot. From wiggle, we see waveform difference because perfect zero-wavelet used in modeling but we don't what the wavelet the real seismic has. Second the model is still not close to real world model yet since this experiment is about consistency test. velocity in the model is not real velocity for the real model, thus amplitude (reflection) will be expected different. Again, the modelled seismic needs to be migrated to reduce mispositioning of seismic events.
Figure 21. Difference of line 1775 between real seismic and modeled
Figure 22. Difference of cross line 3845 between real seismic and modeled
Figure 23. Difference of time slice at 3028ms between real seismic and modeled