THE 3-D DATA

Amoco have collected a complete spectrum of seismic data over the 3-D reservoir survey area. The Amoco Research seismic crew have recently shot a 3-D $\grave{P}\!\acute{P}$ survey, a 2-D $\grave{P}\!\acute{S}$ survey, some 9-component profiles, a multicomponent VSP, an ``AVO'' VSP, a dipole sonic (anisotropy mode), a full suite of logs, and cores. There is a good chance that the 3-D $\grave{P}\!\acute{P}$ survey will be repeated in the near future to monitor a nitrogen injection pilot which is currently underway.

The project is a joint effort between Amoco research geophysicists and production engineers, and the engineers can provide detailed information regarding anisotropic permeability and sorption properties in the coals, constraints on stress-dependent permeability, production history matching tailored specifically to coalbeds, and detailed information regarding brittle failure in coal.

The 3-D $\grave{P}\!\acute{P}$ data are correlated Vibroseis, recorded on frozen ground and packed snow surface conditions. Here are a few acquisition details about the 3-D $\grave{P}\!\acute{P}$ dataset:

• Correlated Vibroseis data.
• Total area covered = 2.6 sq. km.
• Number of shots = 96 per swath * 14 swaths = 1344.
• Traces per shot = slightly less than 600.
• Sample interval = 2 ms.
• Data storage size = 2 GB.
• CMP bin spacing = 7.6 m inline, 15.2 m crossline.
• CMP fold = 20 to 50.
• RMS velocity of principal target = 4.0 km/s
• Depth of target is at 1 km, or 0.5 seconds.
• Dips are shallow everywhere in the area ( $\le$ 5 degrees).

Figure shows a representative receiver line from the 3-D survey, after gaining by a t² factor. The coalbed methane reservoir is the bright reflection at 0.5 seconds. Note the strong AVO character, and the possible polarity reversal (from positive to negative) at the far offsets. The very weak and low frequency basement reflection is barely discernible at 1.3 seconds. There is strong airwave contamination in the lower left corner of the panel, which is severely aliased. Fortunately, it does not cut across the target reflector. However, the Rayleigh waves at about 1.5 km/s and the near-surface trapped mode waves (snow + soil?) at about 750 m/s do somewhat contaminate the near offsets of the target reflector. Additionally, there are some noise bursts on some traces, and some degree of static variation.

Figure shows a muted version of the raw gather. I show this because the mute pattern cuts across the far offsets of the target reflector, destroying a significant portion of that apparent AVO polarity reversal! Originally, before the mute, there is about 40 degrees worth of aperture illumination on the coalbed reservoir. After the mute, there is effectively only about 30-35 degrees remaining. This has an impact on the ability to accurately estimate shear impedance contrasts in the prestack inversion process: the more angle coverage the better the shear impedance contrast estimate reliability. Perhaps some new work can be done on a frequency and dip-dependent mute algorithm to preserve as much of the coal AVO as possible, yet suppress the critical refraction energy.

Figure shows the spectrum of the raw muted data averaged over the receiver line. Figure shows a small experiment in reducing the surface wave noise. An 18 Hz notch filter was applied after examining the data spectrum and correlating an 18 Hz spectral peak to the surface waves. The notched spectrum is displayed in Figure . As seen, the simple notch did a pretty effective job at suppressing the surface waves. However, it also appears to have changed the phase on the coalbed reflector, and introduced a lot of spurious (?) ringiness, perhaps due to the narrowing of bandwidth. A second coal interval at 1.0 seconds seems to have come in clearer though. But, the basement reflector at 1.3 seconds has all but vanished. This is an important reflector to preserve in order to anchor the variations in the upper coalbed level at the final interpretation stage.

Figure shows a $\tau-p$ slant stack of the receiver line. Most of the aliased airwave and low-velocity surface waves are well isolated in this spectrum. In particular, the coal reflection at 0.5 seconds is well separated at a fairly high moveout ray parameter of about 0.1 s/km. This appears to be a good domain in which to try an amplitude-preserving coherent noise suppression by least-squares inverse slant-stacks.

Figure shows a simple NMO velocity semblance spectrum. The coalbed target reflection comes in strong at about 4.0 km/s rms velocity. There appears to be a fairly linear rms velocity gradient starting at about 3.5 km/s near-surface to about 4.5 km/s at the 1.3 second basement reflection. The airwave and surface waves are off the scale of this plot at the low velocities of 0.35 to 1.5 km/s. The velocity domain may be another good place to try a true-amplitude noise suppression, since the reflection energy is at such high rms velocities compared to the coherent noise. This would involve 3-D least-squares hyperbolic inverse stacks if attempted.

Finally, using the linear rms velocity gradient model, I did a quick stack of the receiver line to see a preliminary stack quality. I merely replicated the stacked trace eight times to give it a nice visual effect. The coal reflection at 0.5 seconds stacks in very well, and has a nice Klauder wavelet zerophase shape (this is Vibroseis data recorded on frozen ground). There seems to be coherent reflection energy at 0.7, 0.9, 0.95, 1.0, 1.1 and 1.3 seconds. The basement reflection at 1.3 seconds is very ringy and probably requires a tighter velocity specification, and definitely some redundancy to enhance the signal.