EXPLORATION

Improving the 4D signal in the Valhall gas cloud

 Using a new multiple removal technique, 4D interpretation becomes much clearer. 

Paul J. Hatchell, Peter B. Wills and Catalin Didraga, Shell International Exploration & Production BV.

The Valhall Life-of-Field-Seismic (LoFS) project is the world’s preeminent time-lapse seismic experiment. LoFS consists of a permanent OBC system that allows frequent high-resolution data acquisition for imaging the effects of production. More than six seismic surveys have been acquired since late 2003. The data quality shows an excellent 4D response to production on the flanks of the field, but shallow gas severely attenuates the PP seismic data in the crest.

Long-period multiples dominate the PP data near and in the gas-clouded crestal region. These multiples persist in the time-lapse difference volumes from LoFS surveys because of differences in their arrival times. For some time, we have known that changing tidal levels were part of the reason for these persistent, non-repeating multiples. Conventional multiple techniques can be used, but significant multiples remain.

Chance played a role in uncovering another reason for the remaining multiples. In a separate investigation, we were examining guided wave behavior on OBC data.¹ These waves are affected by changes in water velocity, which in turn is affected by seasonal temperature changes. We would eventually discover that these water velocity changes are responsible for the persistent, remaining multiples.

In this article, we develop a method to identify and remove multiple contamination on stacked data in and near the gas cloud zone. We do this by correlating measured time-lapse timeshifts with known changes in water velocity between surveys. This technique produces improved 4D timeshift maps and interpretations that are consistent with the production data.

BACKGROUND

The shallow-water Valhall field produces from porous chalk reservoirs that compact substantially during primary production. The reservoir shows large 4D amplitude and timeshift signals.

The crestal region of the reservoir accounts for more than half of the historical production. Yet, it is blanketed by a gas cloud, and the PP data are dominated by multiples. This multiple energy contains short- and long-period multiples, but the patterns are dominated by long-period ones. These long-period multiples are found in many locations in the field. In the gas cloud regions, attenuation and mis-stacking have reduced the strength of the primary arrivals such that multiples represent a large percentage of the remaining signal. These multiples do not subtract away between LoFS surveys because of differences in the multiple arrival times, which result from changing tidal levels and water velocities.

4D TIMESHIFT ATTRIBUTES

Time-lapse attributes that work well on the field’s flanks include amplitudes and timeshifts. As shown in Fig. 1, the timeshifts are more robust than the amplitude attributes. Over the crest, the effects of gas cloud attenuation become larger; the reliability of amplitude attributes degrades faster than that of the timeshift attributes.² This makes sense, since the timeshift measurement can be made using large time windows that are excellent at suppressing random noise. In Fig. 1, the two red polygons depict the inner and the outer gas cloud regions, while the black polygons show the producing wells.

Fig. 1. Time-lapse timeshifts between two surveys, RMS amplitudes and normalized RMS amplitudes of the difference between aligned monitor (Survey 5) and baseline (Survey 1). An RMS gate of +/- 40 ms around the top of the reservoir has been used. This figure clearly shows that the timeshifts are more robust than amplitude attributes.

Ideally, we would like to see regular, progressive changes in color intensity on the timeshift maps, i.e., monotonic color changes. Figure 2 shows the top reservoir event at about 2,500 ms of the 4D timeshifts measured between the first LoFS survey (LoFS 1) and successive surveys, up to LoFS 6, beginning with successive baselines.

Polygons on the Baseline 1 maps represent producing wells whose area is proportional to the production during the appropriate time interval. The central crestal region is outlined in purple. Careful examination shows that in the left and right part of each map, monotonic changes in red (positive timeshifts indicating a slowdown in the overburden) increase in a regular fashion as we move from bottom to top across the repeated surveys (LoFS2-6). They also increase monotonically from right to left, across the baselines. However, the timeshifts in the central gas cloud region are generally much larger than those outside the gas cloud; they are not monotonic and do not correlate with production data.

The culprit: Water temperature changes. As a result of work done on guided waves, we found that the different behavior of the timeshifts, particularly in the crestal area, correlate with water-velocity differences between the acquisition times of the different LoFS surveys. These velocity differences give rise to multiples that do not subtract out. Figure 3 shows measured water velocities1 from an analysis of guided modes from four of the LoFS surveys, and by extrapolating this trend, we can predict the approximate water velocities of the remaining two surveys that were not analyzed (LoFS 4 and 5).

Fig. 3. Differences in water temperature create variations in water velocity. These show up in the 4D data. Baseline-monitor (BnMn) pairings make sense of what is happening to the crestal area: Water velocity changes are creating multiples that do not subtract out.

Figure 3 also shows when various baseline and monitor surveys were taken. It was discovered that the crestal area shows a consistent response when water velocity is taken into account. Baseline and monitor pairs B1,M2 and B4,M5 were both taken under similar water temperature conditions. In both cases, the response shows the gas-cloud timeshifts as positive (red), i.e., indicating that the monitor signals arrive later. These are shown on the corresponding panels in Fig. 2.

Fig. 2. Maps of time-lapse timeshift attributes at the top reservoir event between various vintages of the LoFS surveys. Producing areas right and left of the central crest (purple outline) show regular, gradual color changes across repeated surveys that correspond to production. The crestal area appears to exhibit random changes.

Baseline and monitor pairs B2,M3 and B5,M6 were both taken under conditions of different water temperature, such that the gas-cloud timeshifts are negative (blue), as expected. That is, they indicate that the monitor signals arrive sooner, as shown in Fig. 2.

Finally, for the case when the water velocities are the same for both baseline and monitor surveys B1,M4 and B2,M5, the gas-cloud timeshifts are essentially unchanged.

Figure 4 shows a crossplot between the measured time-lapse timeshifts at 1,200 ms and the differences in water velocity between the corresponding survey pairs (yellow squares). There is a clear linear relationship between the anomalous timeshifts and changes in the water velocities between the different surveys. Such an excellent correlation is expected only if multiples are present in the data. The red line in the plot shows a simple model prediction of the timeshift that results for a multiple arrival which passes twice through the 70-m water column.

Fig. 4. Observed timeshifts are linear with changing water velocity. Linearity demonstrates that these are indeed multiples.

Modeling the multiples. As shown in Fig. 5, whereas the baseline surveys have multiples added to primaries, the subsequent monitor surveys will have time-shifted multiples added to their primaries, which will not subtract away across the surveys. These time-shifted multiples can be positive or negative depending on the water velocity at the time of survey. Normally, these multiples have no appreciable effect on the difference volume, because the primaries are strong. But in the crestal area, where the primaries are weak, they can have a pronounced effect.

Fig. 5. Modeling of non-aligned multiples.

Removing multiples. The multiple contaminations can be partially removed from the 4D timeshift data. This involves measuring the correlation between known water velocities and the 4D timeshift measurements, and then subtracting observed correlations from the data. By doing this independently at each sample time, this simple approach can be used to remove multiples from the entire timeshift volume. Figure 6 shows a time-lapse volume before and after removal of the multiples. Note that strong events due to the effects of production persist.

Fig. 6. Survey 1-5 timeshifts after 500 days (top). Survey 1-5 timeshifts after multiple correction (bottom).

Figure 7 shows maps at the top reservoir event after applying this technique to the various surveys, in comparison to Fig. 2. We see that much of the irregular timeshift behavior near the central part of the field has been removed. The resulting timeshift map changes monotonically as we proceed from LoFS1-2 to LoFS1-6, as well as across the baselines. This technique improves our ability to image production-related changes near and under the gas cloud using the PP data.

Fig. 7. Maps of time-lapse timeshift attributes after removing water velocity correlations.

CONCLUSIONS

Multiple energy is present in many locations in the LoFS PP seismic data at Valhall and dominates the time-lapse response in the gas cloud region. A simple technique using known water velocities allowed the identification of these multiples and the partial removal of their impact on time-lapse timeshift measurements.

The above technique is applied post stack. It is only a partial solution to this problem and does not allow us to correct the seismic images. Additional techniques for suppressing the multiples should be considered in this field to improve the images in the gas cloud region. One simple solution might be to change the acquisition of the shooting vessel from shot lines that run parallel to the OBC cables to shot lines that run perpendicular to the cables.

Differences between tides on neighboring lines will help stack out multiple contamination better using the orthogonal shooting direction. Calvert3 taught us an additional means for suppressing multiples by acquiring repeat monitor surveys at different tidal states and deriving a filter to remove the multiples from the time-lapse differences.

ACKNOWLEDGEMENT

The authors thank Shell International E&P and the Valhall partnership (BP Norge, Norske Shell, Total E&P Norge, and Amerada Hess Norge) for permission to publish this article. The results and opinions presented in this paper do not necessarily represent the view of the Valhall partnership.

LITERATURE CITED

¹ Hatchell, P. J., Wills, P. B. and M. Landro, “Analysis of guided modes on permanent OBC data,” 69th MTG. Eur. Assn. Geosci Eng., 2007.
² Hatchell, P. J., Kawar, R. S. and A. A. Savitski, “Integrating 4D seismic, geomechanics, and reservoir simulation in the Valhall oil field,” 67th Mtg. Eur. Assn. Geosci. Eng., C012, 2005.
³ Calvert, R. C., “Insights and methods for 4D reservoir monitoring and characterization,” SEG/EAGE Distinguished Instructor Short Course, 2005.

BIBLIOGRAPHY

Kommedal, J. H., Barkved, O. I. and D. J. Howe, “Initial experience operating a permanent 4C seabed array for reservoir monitoring at Valhall, “ 74th Annual Int’l Mtg.. SEG, Expanded abstracts, pp. 2239-2242.

THE AUTHORS
	Paul J. Hatchell joined Shell in 1989 after receiving his PhD in theoretical physics from the University of Wisconsin. He began his career at the Bellaire Technology Center in Houston and spent seven years researching and developing shear-wave logging technology, rock properties, quantitative seismic amplitude analysis, and 3D AVO applications. Following this, he transferred to Shell’s New Orleans office and spent four years pursuing oil and gas exploration in the Gulf of Mexico. In 2000 Paul joined Shell International E&P in Rijswijk where he is currently a member of the Areal Field Monitoring team. His main focus is the integration of time-lapse seismic data with the reservoir engineering and geomechanical disciplines. Paul has been named Shell’s “Principal Technical Expert” for reservoir monitoring.
	Peter Wills joined Shell in 1987 as a research geophysicist at Shell’s Bellaire Research Center in Houston after graduating with a degree in physics from SUNY, Stony Brook. From 1987 to 1996, he worked on various projects in seismology including geo-tomography, microseismic acquisition and interpretation and seismic imaging. Wills then joined SEPCO New Orleans and worked until 2002 as a seismic interpreter, mainly on subsalt opportunities. In 2002, Wills moved to Rijswijk and continued seismic until 2004, when he joined the time-lapse research team.
	Catalin Didraga earned his PhD in 2004 in theoretical physics from the University of Groningen, the Netherlands. In 2007 he joined the R&D division of Shell International Exploration & Production BV as a member of the Areal Field Monitoring team.