J. P. Lindsey
GX Technology LP
Houston
There are many parallels between exploring the subsurface using seismic reflections and looking at objects below the water's surface.
In the atmosphere, our vision involves only line-of-sight travel from the object to our eye. But when you look at a fish in a pond, water is added to the visual path. If the surface of the water is glassy smooth, we see the fish in its true form and shape, but its location is not where it appears. The apparent position shifts because the water's differing velocity for light propagation changes the direction of travel for that reflecting off the fish.
It gets worse. Suppose we drop a rock in our fish's pond. The surface ripples break the fish's image into pieces. If a brisk enough breeze comes up, we may even fail to recognize the object as a fish through the rough surface.
The same phenomenon is a geophysicist's dilemma. He must "look" at geologic targets (the fish) using sound (light) that travels through rocks whose velocities (refractive indices) differ. What he sees is then distorted to the extent that varying velocities obscure the sharpness of his seismic image.
The most offensive rocks are those closest to us-near the surface-just below the seismic source and receivers.
We acquire redundant seismic data and "process" several traces to get a more noise-free view of the subsurface. But this processing (addition of the redundant data) is ineffective if the reflections being added are not in time register. And that becomes the problem.
In the Texas Panhandle, the Blaine Gyp, just below our shot holes, is leached out by surface water and replaced by lower velocity material. At some places, surface water fails to penetrate to the Blaine, and leaching does not occur. Thus seismic receivers at different positions see reflections from the same geologic boundaries at different times, creating the timing problem.
Or, in Michigan, the problem is caused by glacial till whose constituent material is so varied that velocity variation from one receiver location to the next causes the same timing problem. Almost every land location presents this same generic timing problem, although its cause may vary.
THE ALASKAN SURFACE
In Alaska, this same problem is occasioned by the permafrost-that perpetually frozen material that runs from near the surface to depths of 1,000 ft. Ice sound velocity is about 10,000 fps; water is more like 5,000 fps.
All we need is a sheet of variable-thickness ice at the surface to create our seismic timing problem. It never fails to happen.
There are even cases of supercooled ice reported, with temperatures well below 32 F. These ice velocities can get as high as 20,000 fps for ice around 0 F.
Because of this contrast between water and ice velocities and the prevalence of the permafrost, timing problems are automatically anticipated for every seismic survey in the Alaskan North.
As if this were not enough, more problems with velocity exist much deeper in the subsurface.
Fig. 1 shows a geologic column for the Fish Creek Area at the eastern edge of National Petroleum Reserve-Alaska (NPRA). Above the Pebble Shale and Kingak formations, which contain many rocks of exploration interest on the North Slope, is the Torok formation.
The Torok sits on a shale with high organic content and a large gamma ray response. It is commonly called the Gamma Ray Zone (GRZ). Where dip is present, the lower Torok units depend on frictional forces at the GRZ interface to maintain their integrity.
But the GRZ possesses good lubricating qualities, and as a result, the Torok "fails," breaking into large disconnected chunks that push the GRZ into thick mounds as they slide along the interface under gravitational force.
These failures can occur in a variety of ways, as discussed in an article by Paul Weimer of the Department of Geological Sciences at the University of Colorado.
The mounding of masses of the GRZ creates a highly changing ratio of GRZ to Torok thicknesses.
GRZ shale velocity is about 9,000 fps; the Torok velocity is more like 12,500 fps. And so we have another timing problem for seismic data processors.
The GRZ-Torok problem affects only seismic times for reflections from below the GRZ level. Above this point, the permafrost is the only cause of timing problems. Below the Torok, both the permafrost and the GRZ affect reflection times.
PICTURE OF THE PROBLEM
Fig. 2 is a portion of a seismic line from the Fish Creek Area in the eastern part of NPRA. The portion shown covers 70,000 ft of traverse.
Formations of interest are labeled along the timing scale in the figure. The "wavy" characteristic of all reflections in the sequence below the Torok is evidence of our GRZ and permafrost in action. Above the GRZ, most visibly in the foreset bedding of the Upper Torok, the effects of permafrost alone are seen.
Fig. 3 is a cross section with the same horizontal dimensions as Fig. 2. It is inspired by Fig. 1. No attempt has been made to exactly dimension the depth cross section of Fig. 3 to the seismic picture of Fig. 2. This is of secondary importance since we only wish to study the nature of these two timing problems.
All of the important interfaces seen in Fig. 2 are included in the cross section of Fig. 3. Our goal is to demonstrate our hypotheses about the permafrost and GRZ timing effects by calculating the seismic response of Fig. 3. If we are successful in designing a cross section with timing properties similar to real seismic data, the basis for the design of methods to correct these aberrations is our cross section, or model. This is the investigative, tutorial, and design use of seismic simulation.
Fig. 3 is designed to include permafrost and Torok failure. To properly model any physical feature, its description must be provided in sufficient acoustic detail. This includes the velocity and density of rock units, their size and shapes as expressed by interface geometry, and possibly other descriptors such as how velocities and densities transition vertically and horizontally.
Fig. 3 encodes the permafrost in the simplest possible way: a flat surface, a bottom-depth variation somewhat like old lakes might cause, and a velocity for normal ice temperature. The base is very likely transitional, not allowing the higher-frequency, tightly curved wave fronts to organize discrete reflections. Modeling this is expensive. Since only the effects on deeper rock reflections interest us, this simple encoding is adequate.
The GRZ/Lower Torok interval is shown in Fig. 3 as having dramatic Lower Torok failure, resulting in GRZ buildups that reach 500 ft in thickness. While this is probably a little dramatic, we have some license in a learning model to see what happens under the most dramatic conditions of our geology.
THE SEISMIC RESPONSE
Fig. 4 is the seismic response to the cross section of Fig. 3. Our reason for study is apparent: the timing variations provided by the permafrost and the GRZ buildups.
The evidence of permafrost variation is the "wavy" appearance along any reflector, moving up and down in step with other reflections. This begins high in the time section and extends considerably deeper.
Note that the local changes in time are inverse to the permafrost thickness. Because it is the high velocity, reflection time reduces when permafrost thickens. In other environments, the variable facies might be low velocity. Then we would expect later times for events under thicker units.
The foreset bed responses in the Upper Torok interval, on both the model Fig. 4 and the real seismic in Fig. 2, show these timing anomalies.
We may safely smooth through these responses in interpretation, thus restoring them to their more proper depth form. We may even calculate the change in permafrost thickness using its velocity and that of the unfrozen material below it, given the time difference between our smoothed interpretation and original reflection times.
This information is required to make early processing corrections aimed at removing these anomalies.
The main seismic interest, however, is below the GRZ. The seismic appearance is obviously affected by the permafrost and GRZ timing anomalies. But how important is this to our interpretation?
It's obviously not too critical for structural interpretation because we are not looking for pimples the size of these timing anomalies. But for any interpretation that depends on character analysis of the seismic data, it is very important.
Fig. 5 vividly illustrates this. One common depth point (CDP) gather of traces has been generated and centered on the 43,350 ft position in the model. These are the redundant measurements of reflection properties at this point. They will ultimately be added, with time corrections, to generate one final trace in the seismic section.
The ray path system to the bottommost interface is shown in Fig. 3. The left part of Fig. 5 is a 36 trace gather from this location. The right part of Fig. 5 is this same gather with moveout corrections done about as well as possible.
Here we note the whole point of our discussion: All data elements are not at the same time position. The stack (addition) of these traces is shown in the center of Fig. 5. It is clear that high frequencies in each individual trace have been lost in this sum.
But even more important is the character change. Major differences can be seen between the unstacked data and the stacked trace. No single event of the stack corresponds exactly to any member of the stack. This would be serious for any stratigraphic interpretation.
While this modeling has helped us see what happens when timing errors persist, it is useful for more than interpretation questions.
We may evaluate any proposed processing procedure to cure the problem by applying it to the output from a model such as this. To evaluate a process on real data only is flying blind by comparison.
ADDITIONAL CONSIDERATIONS
There are phenomena on seismic records that cannot be accurately studied with normal-incidence ray tracing as used here.
The seismic sources and receivers simulated in this type of modeling are coincident on the surface. In the real world, seismic data are acquired by use of a wide range of source-receiver offset distances. Stacking groups of traces having differing offsets can cause phenomena not reproducible with normal incident modeling.
A more complete study of the effects of permafrost and the GRZ on actual seismic data could be made using offset ray tracing, as illustrated by the 36 trace gather of Fig. 5. This should be considered in any modeling exercise.
Vertically aligned timing anomalies are expressions of significant lateral velocity variations above the anomaly. If these variations are of short lateral duration, they can be observed and eliminated in interpretation. However, they will affect the seismic processor's ability to get a good quality stack if they are not corrected.
If the velocity variations occur over greater lateral distances, their timing effects can become confused with structural timing, and errors in interpretation may occur. This is where the strength of modeling pays off. It can be used to define the problem, design its solution, test the resulting process, and guide the interpretation of the results.
Copyright 1992 Oil & Gas Journal. All Rights Reserved.
