Geotrace Technologies Inc. For FORCE Seminar
A Robust Approach to Lambda-Rho Mu-Rho Generation
By
Sean McQuaid, Roman Bobolecki, Richard Dolman (LFP Division)
Abstract
Most processing departments do a good job in providing an objective image of the subsurface from seismic data. However, the reservoir geophysicist is often narrowly focussed and intent on extracting the maximum information from a narrow time interval. Under these circumstances, a number of processing issues require revisiting after the first stage of processing. In order of importance, they are:
· Offset Phase Variation
· Offset to Angle Conversion
· Offset Scaling
· Offset Time Variation
· Offset Bandwidth Variation
Offset Phase Variation
Processing is usually performed at a consistent phase. Although no processes are applied that would give a different phase on the far offset compared to the near offset, sometimes it is observed that a peak on the near section might look like a trough peak on the far. A single boundary might therefore, appear to show an extra event on the fars.
This is corrected by cross-correlating the nears with the fars and inspecting the phase of that cross-correlation (which relates to the difference of phase between the near and far traces). The correction is always back to the phase of the nears. In this approach Phase variation is defined in conjunction with the determination of Timing differences.
Figure 1 shows a cross-plot of all the phase and time shift differences within a 3D survey for a particular time interval. Each dot represents a single cross-correlation between a near and far trace, with time differences likely to be caused by residual NMO.
Negative correlation between the time shift and the phase shift is evident. As theory predicts, the slope of the distribution is defined by its central frequency, the extent of the distribution along it’s major axis by the bandwidth of the data and the distribution along it’s minor axis by the signal/noise ratio.
When defining a systematic phase correction, residual timing errors should be ignored. In this case there is an 8° correction.
Offset to Angle Conversion
The equations used in AVO theory relate amplitude changes with angle, to shear and compressional wave velocities. Pre-stack data invariably come sorted in offset. Mistakes in conversion from offset to angle will result in a compromised AVO response.
A simplified conversion from offset to angle, is often used resulting in iso-angles which, when overlaid on an offset sorted CDP, monotonically increase in time and offset. In reality, the incident angle for a deeper sample on a constant offset trace can be smaller. Figure 2 shows an accurate Offset to Angle conversion, developed by ray tracing through a sophisticated velocity model. Here, the coloured lines show the relationship between constant incidence angles and offset. Dramatic velocity changes result in angle mutes not always increasing in offset with depth, as shown at 3.1 seconds.
Offset Scaling
Any processing that changes the amplitudes differently for different offsets/angles will mask true AVO changes.
For example, a spherical divergence correction preferentially boosts amplitudes of events with greater arrival times. If this is used inappropriately, once NMO is applied an AVO effect due to spherical divergence correction and not lithology will be observed. Processors attempt to prevent this, but confidence is needed that processing induced AVO anomalies are absent.
Vp, Vs, and Density logs from wells and horizons are used to produce synthetic model CDPs. The average amplitudes of the real data are divided by the synthetic average amplitudes, to create a “scaling CDP" for every CDP. Figure 3 presents an example "scaling CDP" in colour with the real CDP as a wiggle overlay. The purple colour indicates the need to multiply the real seismic by 80,000 to convert to reflectivity.
Any systematic changes in scaling with offset, time or location can therefore, be identified and can be considered for a scaling correction strategy.
Offset time variation
Even after pre-stack depth migration, it is unlikely that all pre-stack primaries will be perfectly flat. If they are not, AVO anomalies can appear that are purely related to residual NMO.
Peak-trough matching and cross-correlation techniques are used to create volumes of time shifts required to align near and far angle volumes.
In the top image in Figure 4, the top left arrow, shows an example of misalignment, with the near trace trough (wiggle) associated with a peak/trough (in colour: Red-White-Blue) on the far stack, owing to residual moveout.
The bottom section shows that the same trough on the nears now relates to a trough (Blue) on the fars after making the timing corrections.
Offset Bandwidth Variation
Far offsets have a very narrow bandwidth compared to near offsets, predominately owing to NMO stretch.
Proprietary inversion techniques have been developed that provide seismic derived information half an octave higher than is conventionally observed. Resulting Lambda-Rho and Mu-Rho volumes are of higher resolution, allowing anomalies to be mapped in more detail.
The final AVO product
The careful application of the process described above is essential for reliable AVO products.
Figure 5 shows a perspective view of a sand channel that has a low Lambda-Rho value compared to its background. The value of Lambda-rho will change according to the fluid fill. The blue channel is gas filled. The left hand channel is predominantly blue up-flank and green (slightly higher lambda-rho) downflank. This may indicate oil.
In conclusion
This methodology specifically addresses the main concerns in AVO processing and enables more reliable AVO products to be generated.