Abstract Although global reserves figures continue to rise, so too does consumption and population. At some stage, a fundamental resource limit will be reached, and there will be a supply and demand discontinuity. Since hydrocarbon recovery from existing reservoirs is generally low, the most significant impact on supply is likely to be made by increased oil recovery methods. In reservoirs, information between wells is scarce, and seismic monitoring techniques have tremendous potential to provide such lateral data provided there is a clear link between seismic observables and reservoir variables such as fluids, pressure or temperature.

Abstract This section lists and "flags" those reservoir properties that may change with time and result in seismic differences. It also gives insight into the typical extent of the changes that may be experienced. A later section covers predicting possible seismic changes. During the hydrocarbon extraction process, those primary reservoir properties that change with time include: Secondary effects of hydrocarbon extraction include compaction, porosity, density, overburden stress, fracturing and chemical changes.

Abstract This section covers predicting the seismic changes that might be observed as a result of changes in reservoir properties over time. The starting points are the simple connections between rock properties and the (zero offset) seismic reflection event, relating 1) the reflection coefficient to the acoustic impedance (AI) change at a boundary: These must be related to primary reservoir changes (or pore fluid composition, pressure, and temperature) and to secondary effects, so we try to quantify the effect of the primary changes on reservoir rock bulk density, compressional wave velocity, and on Poisson's ratio (the ratio of the fractional transverse contraction to the fractional longitudinal extension as a volume of material is stretched). In particular, we would like to be able to predict the seismic response changes that might then occur, as in the cartoon in Fig. 3.2, both at zero offset and at larger angles of incidence. Bulk density (Fig. 3.3) is the sum of fluid density times porosity plus the matrix density effect. Clearly, in low porosity reservoir rocks, fluid changes from brine through oil to gas will make little difference to the bulk density, so one of the first success factor requirements is the presence of a high porosity reservoir rock. As reservoir pressure depletes, the overburden stress causes compaction with reduced reservoir porosity and increased density. The higher the initial porosity, the greater the impact of pressure depletion (Fig. 3.4). A pressure increase, however, is unlikely to produce a reverse effect. The effect of temperature on

Abstract This section expands on some of the items listed in Section 2C. It is intended to impart a healthy suspicion of the effects of data processing parameters. These effects can be much reduced by well matching/wavelet extraction methods, but this should not be assumed. In differencing operations noise becomes much more of an issue! Time-zero corrections and statics applications: The data must be repeatable in time. Just a millisecond of time shift allows root mean square (rms) amplitude differences of around 18% in a simple test. Fig. 4.A.1 shows a "final" processed migrated section from a 3D data volume. In the lower panel of the figure is a subtraction produced from the same section, time shifted from itself by 1 ms. To show a more extreme example, consider the 4-ms shift shown in Fig. 4.A. 2. This produced a subtractive rms amplitude of 71% of the original data. We can conclude that time matching of seismic data is critical before its use as time-lapse data. Since many of our "final" sections exist at a sample interval of 4 ms, we have to ensure that any software used to compare several 3D data cubes has an interpolated time shift capability. Acquisition equipment and data processing time delays may easily impart a fraction of the final sampling interval Mute applications: An early step in a processing sequence is often a first-break mute. This will be picked empirically from trial parameter tests and examination of the data. The effect of a small change

Abstract As illustrated by the foregoing material, data from two or more time-lapse surveys are quite likely to have repeatability problems. One question that therefore arises is: how successfully or justifiably can we "force" a match between them? After all, 99% of the earth volume covered by the survey will not have changed perceptibly between the two surveys. In this section we investigate a few guiding examples of attempts to do this. Fig. 5.1 shows a dip line from two co-positioned 3D data volumes from 1985 and 1995 surveys–a typical "legacy" data situation. The oil-water contact is initially at 1.9 s on the 1985 data, and the column has moved about 50 m upwards during the 10-year interval. It also shows the difference section, first that obtained with just a single "global" amplitude matching scalar, and without any trace "matching" of any kind, although the two data sets have been processed with time-lapse intentions. Clearly, the subtraction has not been especially successful, although there is a strong event in the region of the oil-water contact.

Abstract Now that we have worked through enough background, and with the experience also of looking at several case histories, we are ready for a summary session on "how to do it" and some final points and "lessons learned." (A) Work the rock physics. Establish the link between reservoir properties and seismic observables using logs, core and fluid data, plus VSP data if available. (B) Work the reservoir simulator. Determine the changes we would expect due to production–the new fluid distributions and pressure changes. (C) Work the seismic modelling to predict the changes in the seismic observables. Generate synthetic seismic data and assess the magnitude and character of the changes that might be visible. Assess the need for any special processing or seismic attribute studies. Study the resolution. (D) Work the economics to estimate where and how the time-lapse data adds value. (E) Design the baseline seismic survey (or the repeat). (F) Work the earlier or later seismic data–it may be possible to extract more from it. (G) Interpretation and modelling of time-lapse seismic relies on good well matching and wavelet extraction. If this is not possible, the error bars on the interpretation will be much larger. (H) If there is no base reservoir or OWC event, then events below the reservoir may still allow a good indication of change via observable time delays. (I) Manage the expectations. Results are more likely to be qualitative than quantitative. Don't over-hype. (A) Rock Physics–Some key factors: Pore fluid affects the overall compressibility

Abstract I am indebted to many people who have supplied their data, their time for discussion, their encouragement, inspiration (some unknowingly), and their assistance in many ways. Some of these kind people are listed here, and for completeness I also include those who persuaded me (maybe maneuvered is a better word) into preparing this presentation.

Abstract An experiment testing In-situ combustion with oxygen as an enhanced oil recovery process was run at the Holt Sand Unit in Montague County, Texas, from 1981 to 1983. This report integrates the interpretation of geological and geophysical data collected by the Exploration and Production Research Department of ARCO Oil and Gas Company during the experiment. The report Illustrates monitoring of the burn in time using 3-D seismic data and describes the factors in the detailed geology of the reservoir which controlled the propagation of the burn. The most important feature of a reservoir involved in a flooding type recovery program is formation permeability. The Holt Sandstone was found to contain a complex array of vertical and horizontal barriers and enhancements to permeability. The reservoir complexity required that a proper description of permeability include not only measured permeabilities from core samples, but also a detailed study of the depositionai history and post-deposltional alteration. The original permeability of the Holt, which was controlled by depositionai processes, has been overprinted by natural flushing, secondary cementation, and fracturing. A geologic model of reservoir-scale permeability is shown to be useful in interpreting the details of the combustion process.

Abstract This article describes a successful 4-D seismic pilot project in Duri Field, Indonesia, the site of the world's largest steamflood. Two baseline and six monitor 3-D seismic surveys were recorded over the same steam injection pattern between 1992 and 1995. Seismic analysis, thermal simulation, and acoustic modeling show the engineering relevance of these data and provide new insights into the seismic response of reservoirs during active steamflooding. Analyses of the data demonstrate that, for the Duri steamflood, the horizontal and vertical distribution of steam can be tracked over time. Based on these results, an economic feasibility study was conducted to investigate potential future application of 4-D seismic at Duri. Decision analysis tools then determined the impact of seismic monitoring for potential reservoir management and steamflood development. The study concluded that a large increase in the value of the steamflood could be realized if seismic monitoring were applied on a large scale. The pilot study led to the implementation of a large scale time-lapse 3-D seismic monitoring program, discussed in the companion article "Time-Lapse monitoring of the Duri steamflood: Application

Abstract A technique is illustrated which uses a single 3D seismic monitor survey along with multiple seismic attributes in conjuncture with an offset baseline survey to discriminate heated reservoir from unhealed reservoir. 3D seismic monitoring has definitively mapped steam heated regions of the reservoir at two separate geographic locations encompassing more than 150 wells in Cold Lake. Based on these maps infill wells have been drilled into those regions of the reservoir which have remained unhealed after ten years of steam injection.

Abstract Time-Lapse seismic analysis has been applied to two 3-D seismic surveys acquired over the Central North Sea hulmar Held -- a pre-production survey shot in 1977, reprocessed in 1987, and a 1992 survey. The Upper Jurassic reservoirs in the field have been under production since 1982. Water is the main drive mechanism, supported by flank injection. Although the field is currently at over yi)7o water cut, there are inrm drilling opportunities. hetropnysicai analyses for huimar indicate that water replacing on win result in an increase in seismic impedance. In addition, a pressure decline of about 1000 psi during the time between the two seismic surveys will result in a further impedance increase. These impedance changes are observed between the two seismic surveys. In order to overcome inherent differences in the seismic data due to acquisition and processing differences, the data are equalized and then inverted to obtain impedance which is then averaged between the top of the reservoir and the position of the original oil-water contact. Differences in averaged impedance between the 1977 and 1992 surveys clearly show the effects of water influx and pressure decline. The changes observed in the seismic data are overall consistent with predictions obtained from a full-field, history-matched flow simulation as well as production data. Differences in details may suggest areas of bypassed oil. However, data quality is not sufficient to serve as the sole basis for drilling decisions.

Abstract The Reservoir Characterization Project (RCP) is an industry sponsored consortium whose mission Is to develop and apply 3-D and 4-D ("time-lapse"), 3-C seismology and associated technologies to improve reservoir performance and hydrocarbon recovery while reducing environmental impact. RCP-Phase VI is the multidisciplinary, 4-D, 3-C study of a pilot C02 "huff-n-puff" injection project in Vacuum field, a shallow shelf carbonate reservoir located on the Northwestern Shelf of the Permian Basin, Southeastern New Mexico. The C02 "huff-n-puff" occurred in well CVU-97 at Vacuum Field. The initial 3-D, 3-C survey was acquired from October 28 through November 13, 1995, prior to the C02 injection which occurred from November 13-December 8. The "soak" period extended from December 8 through December 28, after which CVU-97 was returned to production. The second 3-D, 3-C survey was acquired from December 21 to December 26 during the "soak" period.

Abstract From 1987 to 1990 the Alberta Research Council and Western Atlas conducted a major research program to assess the feasibility of monitoring hydrocarbon miscible floods in carbonate reservoirs with time-lapse seismic techniques. The results of this study, based on ultrasonic velocity measurements, indicated that saturation-induced velocity contrasts are significantly larger than Gassmann predictions in the majority of carbonate reservoir types. Seismic modelling, based on these results, suggested that injected gas should be seismically detectable in many carbonate reservoirs. While these results are encouraging there are major uncertainties in relating the velocity changes observed at the high frequency laboratory scale to the low frequency seismic scale. We have recently conducted several theoretical and experimental investigations to relate the velocity changes at core scale to the seismic scale. These newer findings tend to suggest that it should be possible to track the movement of injected gas in many carbonate fields using time-lapse seismic methods. This conclusion was indirectly supported by a detailed modelling study and a 2D test line acquired over a solvent injection well during 1987. Continued injection during the past 10 years has increased the thickness of the gas bank from 10 m. to 30 m. This test line has been recently recorded again, in March 1997, to test these concepts.

Abstract The experiment: In this case history, the French gas supply company Gaz de France planned to store gas by injection into a water-bearing sandstone reservoir in a faulted anticlinal structure at a depth of about 900m in the Paris Basin. It consists of a combination of sandstone channels of excellent reservoir quality with an average thickness of 20m. An R2 reservoir exists below Rl and is made up of complex communicating channels. Two tops were identified, and wells CE12 and CE112 were drilled for the first gas injections. Such fluid substitution induces a significant change in Vp and density, which suggests that the gas bubble could be detected with seismic data.

Abstract Time-lapse seismic is sometimes referred to as 4-D seismic with calendar time between successive 3-D surveys as the fourth dimension with the objective of monitoring depletion and bypassed pay by acoustic changes due to fluid production. But what if no reference 3-D survey was acquired before production started? Production began in many of today's fields before 3-D seismic was commonly available, but old 2-D seismic is almost always available. The ability to produce accurate difference sections between such disparate surveys expands the potential applicability of time-lapse seismic.