Bandlimited impedance inversion: using well logs to fill low frequency information in a non-homogenous model

Heather J.E. Lloyd and Gary F. Margrave


An acoustic bandlimited impedance inversion study was done to compare the results obtained when the baseline log; or alternatively, a monitor log are used to provide the low frequency information for the inversion. The model was created using the baseline log for the regional trend and placing the fluid substituted (monitor) log at the injection site in the center of the section. These wells were then interpolated along a geological section with a 2.8 degree regional dip. This impedance section was then converted to reflectivity and convolved with a zero phase, Ormsby wavelet to create a normal incident synthetic data set. The frequency spectrum of the logs and synthetic data were compared to find a suitable low frequency cut-off value. The low cut-off value of 4.5 Hz was chosen as frequencies higher than this were found to produce low frequency smearing across the section. The inversions were carried out using this low frequency cut-off for the logs and the high frequency cut-off for the seismic data was set at 85 Hz. Figure 1 shows the result for the monitor impedance inversion. The baseline inversion had a mean error of 28% at the injection site where the monitor inversion had an error of 23%. Both inversions had a regional error of ±10% when compared to the true impedance given by the model. The cross-correlation between the seismic data and the synthetic seismogram for the monitor inversion case was higher, and the sum of the error between the true impedance and the monitor inversion was slightly lower when compared with the baseline inversion. This shows that the monitor inversion is slightly better than the baseline inversion. It was found that more testing is required, using different models and acquisition geometries to determine which log is better to use in time-lapse studies. The findings of this paper suggest that possibly the best results can be obtained when the baseline log is used for the regional area and the monitor log is used for the injection area.

View full article as PDF (0.51 Mb)