John W. Fairborn

Characterization of a steamed oil reservoir using cross‐well seismology

Cross‐well seismic tomography is developing into an important tool for reservoir management, and within the last few years there have been some notable advances in out understanding its imaging capability. Tomography has evolved from its roots in medicine to encompass a broad range of geophysical applications (which include earthquake hypocenter location, mining, and nuclear waste disposal) in addition to oil field operations. The fundamental technical requirements for the latter have been demonstrated. High‐frequency seismic waves capable of traveling long interwell distances can be generated without damaging the borehole, and mathematical inversion techniques can give reliable images as long as problems associated with nonuniform and incomplete sampling are handled correctly.

Identifying stratigraphic units by seismic patterns

Decomposing the seismic data into classifiable patterns is only the first, and perhaps easier step. We have observed that the se segments can be grouped and displayed as false color images in a way that can be related to seismic facies units. These images can also bring out subtle waveform changes and discontinuities that are easily overlooked on a conventional gray-scale image, while at the same time preserving the essential character of the seismic data. The second and ongoing task, therefore, is to quantify the fine structure (or texture) in the seismic data, relate them to mapable stratigraphic units and separate them from structural features. Such capability would allow us to highlight different stratig raphic facies in false color images in order to peruse very large 3-d images and rapidly focus on regions of interest. Finer grained patterns within the regions of interest could then be calculated. A longer term goal involves supervised searches for seismic patterns associated with know reservoir facies, where the patterns could be determined from the analysis of synthetic seismograms computed from well log information. We show two examples of these applications. Waveform effects from fluid injection: The first example is a simulation of what might be recorded in the second survey of a 4-D program. Figure 1 shows a colorcoded velocity model created by interpolating between 2 sonic logs. The velocity within a 200-by-50 foot rectangular layer was reduced 2 percent to simulate the effects of CO-2 or steam injection. This layer can barely be seen on the Figure; it is loc ated between 400 and 600 feet at a depth between 1500 and 1550 ft. One dimensional, synthetic seismic traces were computed every 10 feet along this section, assuming a density of 1.0 and using a n technique introduced by Goupillaud in 1961. Two sets of synthetic tr aces were computed: one with and one without the reduced velocity layer. Random noise with the same frequency bandwidth as the signal and two-thirds the rms amplitude was added to the traces. The difference between the seismograms computed with and wit hout the reduced velocity layer is shown on Figure 2. The random noise effectively hides the effects of the low-velocity zone. The two seismic section were then decomposed into waveform segments based upon amplitude variations along both the time and horizontal axes. The images shown in Figure 3 and 4 use a color palette that gives the broadest color range to the least frequent patterns. In this way the slight change in reflection amplitude, which outlines the path of fluids injected into the reservoir, can be seen. The change is difficult to see and has been highlighted with an enclosing rectangle. Nevertheless, it is slightly easier to see these fluid effects on the trace patterns than on the velocity model. Correlating reservoirs and seismic patterns The second example shows how seismic and well log measurements can be searched together in order to find unique patterns associated with known reservoirs. These patterns may not necessarily be the same; they only need to stand out against the background

McKittrick Cross-Well Seismology Project: Part I. Data Acquisition and Tomographic Imaging

Summary A cross-well seismic tomography survey was conducted jointly between Chevron anh Teiaco in ihe Texaco portion of the McKittrick oilfield. The McKittrick oil field is shallow, less than 1300 feet, and contains low gravity oil which is perched above the water table. The formation consists of unconsolidated sands, conglomerates and diatomite. The P- and S-wave velocities are slow, sometimes very slow, and the sediments are highly attenuating. A downhole hydraulic vibrator developed at Chevron was used as the seismic source. The primary objective of the survey was to map a sealing fault that separates saturated from unsaturated parts of a thick oilsand. A second objective was to evaluate the performance of the downhole vibrator. The source depths ranged from 200 to 920 feet with mostly a 20 feet vertical spacing. There were four receiver wells, two within 100 feet of the source well, one 3 15 feet from the source well, and the fourth 467 feet from the source well. The vertical spacing between each receiver was 20 feet in the far wells and 10 feet in the near wells. The cross-well data are of generally good quality, and the vibrator operated without problems for over 3ooO sweeps. Both P and Shear wave direct arrivals are clearly seen as well as a number of reflections. The latter can be used to map bed boundaries and apply constraints to the tomogram inversion. The tomogram presented in this paper was computed using an ART algorithm. The sealing fault and saturated zone can be identified by a sharp P-wave velocity contrast, and the velocity distribution provides a detailed picture of the geologic structure between the wells.

McKittrick Cross-Well Seismology Project: Part II. Tomographic Processing And Interpretation

A modified SIRT algorithm was used to invert crosswell seismic data acquired in the McKittrick Field, California. The resulting tomogram successfully imaged a complex fault system associated with the McKittrick Thrust. The fault-controlled reservoir was delineat,ed with a resolution of 40 ft [12.2 m] providing the geologists and production engineers with detailed geology and reservoir heterogeneity informa.tion previously unavailable. The McKittrick Thrust system consists of a main thrust and numerous subthrusts. The tomogram suggests that, a fault plane associated with one of the subthrusts is actually flatter than previously believed. We cross-checked the tomogram-derived interpretation for the McKittrick Thrust system with kinematic structural modeling (balanced cross section method) and found that, the McKittrick Thrust and its associated subthrusts can be classified as an imbricated fault-bend fold structure. Oil saturation for the reservoir was also estirnated from the seismic tornogram by exploiting an empirically derived relationship among oil saturation, temperature, and seismic p-wave velocity. The estimated oil saturation agreed with the average oil saturation values measured in the laboratory.

Imaging of thin beds using advanced borehole seismology

The high cost of data acquisition and the limitations of energy sources have impeded the widespread use of borehole seismic techniques. While piezoelectric and air‐gun sources continue to improve, these fluid‐coupled sources have severe problems associated with generation of tube waves. Additionally, the piezoelectric sources have a narrow frequency bandwidth which peaks above 1 kHz, thus limiting their application to high-Q sediments. Consequently there is a need to complement the fluid‐coupled sources with one which generates lower frequencies and which, by clamping to the borehole wall and thus achieving more efficient coupling of energy into the formation, will propagate energy to a greater range and minimize tube waves.

Unstacked migrated traces for AVO and velocity analysis

There are a number of reasons why it would be nice to directly map reflections on a time trace to their image point in depth without all the trace mixing required by conventional migration. Such one-to-one mapping would preserve the shot and receiver coordinates of the trace, thereby allowing reflections to be modeled within the framework of the data.

AVO Inversion On Depth Migrated Traces

It is generally recognized that avo inversion is better conducted on depth migrated traces, although it is not often done because of computational costs and the fact that it may provide only marginal improvement in regions of low dip and small lateral velocity gradients. However, there are instances where it is necessary, in which case multiple common-offset migrations are undertaken. Aside from being computationally intensive, common offset migration is a trace mixing process where amplitudes can be distorted due to unattenuated migration noise and an inaccurate migration velocity. Careful attention must also be paid to the migration operator weights. Described in this paper is an alternate procedure where a reflection from a single time trace is mapped to its specular reflection point on a single unstacked migrated trace. This gives a one-to-one correspondence between a reflection on a raw input trace and its image. The shot and receiver locations are preserved and the true reflection point is known. The procedure begins with Kirchhoff prestack depth migration where the output from each trace is individually stored at selected image points. Most of the energy on these traces will be migration operator noise; however, the specular reflections (if any) can be recognized by a straightforward, but manual, process that will be elaborated upon later. Because manual intervention is required to pick the specular reflection, this process is not for reconnaissance. Presumably an efficient program for finding avo anomalies has already been run and some anomalies have been singled out for further analysis. Depth migration need only be run within the anomalous region so that the entire process is not as computationally intensive as one might expect. On the other hand, separating the specular reflections from the migration operator noise is manpower intensive and at first glance may seem somewhat tortuous. A good trace sorting program is required. The final result is a set of unstacked migrated traces, sorted by common image point and offset, and with known shot and receiver locations. Moreover the reflection of interest is known to be a specular reflection, not just the limb of the migration operator impulse response, which without the benefit of stacking would be unattenuated. Rays can then be traced to the reflection point, allowing both the incident angle and a geometric spreading factor to be calculated. After correcting for geometric spreading the trace amplitudes and incident angles can be used to compute the coefficients of any one of the common avo functions This procedure provides an opportunity to combine PS and PP avo inversion with a minimum amount of migration effort and no ambiguity with respect to their common reflection points. Misalignment of reflector depths on the common image traces can also be a quick way to evaluate the migration velocity.

The Use of Cross Well Seismology to Characterize and Monitor a Steamed Oil Reservoir

Publisher Summary Cross well seismic tomography is developing into an important tool for reservoir management and within the last few years there have been some notable advances in our understanding of its imaging capability. Tomography has evolved from its roots in medicine to encompass a broad range of geophysical applications, which include earthquake hypocenter location, mining, nuclear waste disposal and oil field operations. The fundamental requirements for the technology have been demonstrated. High frequency seismic waves capable of traveling long interwell distances can be generated without damaging the borehole, and mathematical inversion techniques, such as ART (Algebraic Reconstruction Technique) and SIRT (Simultaneous Reconstruction Technique), can give reliable images as long as the problems associated with nonuniform and incomplete sampling are handled correctly.