Modeling Result - 25Hz Zero-Phase Ricker Wavelet

Synthesized seismogram was computed based 20ft by 20ft by 20ft cell space of the velocity model.

syndata88 is recorded seismogram at the surface (zero-depth) of the model, it has dimension 207x330x600. Total time recorded 600ms. Since source was placed at the surface, we have to remove direct wave from received seismogram, so we need to compute only direct wave for only source function length, in this case 86 time steps. The recorded direct wave is stored in file fsrc88. The direct wave has to be removed from the recorded seismogram in order to compare with real seismic data.

Since seismogram is in 20ft by 20ft space in X and Y, not 41ft by 65ft space in real seismic data, also sampling interval for seismogram is 1ms instead of 4ms of real seismic data, we have to resample seismogram in time and space. Time can be easily by desampling, i.e every 4 samples, pick one sample, this will not have alias problem since wavelet frequency is 25Hz, Nyquist frequency of 4ms sampling is 125Hz. In X and Y, we used the 4D module volume interpolater which does 3D interpolation by taking account of seismic event dipping angle. We generate two new 4D formatted seimogram volume which are stored in files in directory

Here st88_seismogram_20x20.fld is the seismogram with time sampling rate 4ms now, but spatial sampling is still 20ft by 20ft. st88_seismogram_41x65.fld is the seismogram which is spatially now 41ft trace interval and 65ft line interval. Therefore this volume is now comparable to real seismic data in time and space. All figures showing below is timberwolf 1988 model.

Figure 1 is the source function which spans about 128 ms. But we cut some of samples off in the front and tail because they are very small value. Figure 2 is the frequency spectrum of the wavelet shown in Figure 1.

Figure 3 shows frequency spectra and histograms of real seimsic data and synthesized seismic data before global frequency and amplitude normalization. Figure 4 is the spectra and histograms after global normalization-bandpass and global scaling. green color represents synthesized seismic data, and red color represents real seismic data. First we see frequency spetrum is dramatically different, modeled seismic has dominant frequency around 25Hz, but principle frequency of real seismic data is about 16Hz. Amplitude level is different since we do not calibrate the source amplitude. Here we simply multiply each data point of seismogram by a factor 10.7477. After global scaling, we can see histogram of real seismic data and modeled seismic are very close (Figure 4). From figure 4, we use bandpass bandwidth 0-50Hz instead of suggested 0-30Hz by the program, since dominant frequencies of real seismic and synthesized seismogram are very different, we can not fix frequency by bandpassing here. So only way to solve this difference is generating seismogram by using 15Hz principal frequency.

Although the dominant frequencies are different, it is still worth of taking a look modeled result with real seismic by comparing them. We are showing a series of images here.

Figure 5 shows 3 inlines extracted from real seismic volume and modeled seismic volume.

Figure 6 shows 3 cross lines and Figure 7 shows 4 time slices extracted.

Figure 7 shows 3 time slices extracted.

From these 3 time slices, we can see great correspondance between real seismic and generated seismogram, but when we move to deeper, such as time slice 3124ms and 3482ms, we will see seismic features off, this off can be seen from the inline and cross line section shown in Figure 5 and 6, we basically see seismic event dipping angles are are lower on modeled seismic because synthsized seismogram is close tozero-offset post-stacked seismic data, the modeled seismogram needs to be migrated in this case definitely in order to compare with real seismic data.

To show comparison between real seismic volume and synthesized seismogram, we furthur display seismic in overlayed wiggle form. The wiggle plot will clearly show difference between them.

Black color is modeled seismogram and red color is real seismic. We can clearly see frequency difference and time difference. Therefore migration definitely needs for modeled seismogram.

From wiggle plot, generally we can see good match at the shallow part but not the deaper part down to the reservoirs. One reason of course is the wavelet is different, another thing is time different because real seismic is migrated data but not modeled one.

Also the boundary reflection still exists although it is weeker comparing to reflection signal. Boundary reflection contaminates modeled seismogram on the West side of model since almost every reflector in the model is dipping to west. This causes boundary reflection dipping to East which is opposite to the reflection events. Therefore dipping filter can be used to remove the boundary reflection. Figure 10 shows one example for line 1817.

Based on observation we made above by comparing synthesized seismic with real seismic, we are not going to comapre modeled seismic with real seismic of 1994 Western survey. We will concentrate on 15Hz modeling result instead in next section of this page.