Azimutal Avo

P r a c t i c a l a p p l i c at i o n s o f a n i s o t r o p y

750 The Leading Edge July 2011

SPECIAL SECTION: P r a c t i c a l a p p l i c a t i o n s o f a n i s o t r o p y


The ability to identify fracture clusters and corridors and their prevalent direction within many carbonates and

unconventional shale gas/tight gas reservoirs may have a significant impact on field development planning as well as on the placement of individual wells. We believe seismic fracture prediction provides the best opportunity to identify the spatial distribution of fracture corridors, but the reliability of seismic fracture detection technology is constantly being questioned. The criticism results from the degree to which the acquisition footprint, random and coherent noise in the seismic data, and near-surface/overburden issues affect extracted seismic “fracture” attributes. Therefore, a key issue is the separation of artifacts caused by the acquisition footprint and near-surface or overburden anisotropy/structural variations from the anomalies caused by the presence of fractures.

Seismic fracture prediction is based on the fact that, in fractured rocks, seismic velocity varies with the direction of wave propagation relative to the fracture orientation (azi-muthal anisotropy), which causes seismic amplitude to vary with azimuth (azimuthal AVO or AVOAZ). The theory of wave propagation in fractured rocks is well documented, but many issues remain to be resolved before this technology can be used routinely. These issues include overcoming problems related to acquisition and overburden effects and determining optimal data-processing and signal-enhancement methods. In this paper, we apply a target-oriented AVOAZ processing workflow, which includes data conditioning, suppression of the acquisition footprint, and mitigation of near-surface and overburden effects to extract fracture-anisotropy attributes in a carbonate oil field in United Arab Emirates (UAE).

In 2001, 1500 km2 of wide-angle 2C (hydrophone and geophone) OBC data was acquired in a carbonate field. The targets are lower Cretaceous shallow marine carbonates in al-ternating tight and porous sequences. The average porosity is 25% with permeability of 10–100 md. The total thickness of the zone of interest is approximately 106 m (350 ft) with the thickest reservoir being 50 m (160 ft) (Figure 1). The data were processed in 2002 and 2006, and reprocessed in 2009 with a processing flow designed to better image the field and mitigate multiples and noise (Reilly et al., 2009). This study, focusing on a subset of 3D prestack data (6.5 × 9 km2), was intended to assess whether the reprocessed data set could be used for fracture detection.

The OBC survey has coverage in excess of 300 fold within 12.5 m × 25 m bins (CMP spacing was 12.5 m in the inline direction and 25 m in the crossline direction). At the primary reservoir level at depth in the range of 2100–2400 m (6800–7800 ft), the survey provides full azimuths with a maximum

ENRU LIU, GREGG ZELEWSKI, CHIH-PING LU, and JOSEPH M. REILLY, Exxon Mobil Upsream Research CompanyZYGMOUNT J. SHEVCHEK, ZADCO

offset of 4500 m (14,700 ft). The acquisition was orthogonal, with source lines (N27°E) running perpendicular to receiver lines (N117°E). The receiver-cable spacing was 300 m. The regional structure is relatively flat with dips of less than 3°, a favorable condition for performing azimuthal seismic-attri-bute analysis. The unmigrated data were used in this study.

Data preparation and target-oriented workflowIn practice, two methods are often implemented to extract anisotropy information: one is to use velocity and traveltime and the other is to use amplitude. The first method inverts the azimuthal variation in interval traveltime (dt) or per-forms azimuthal NMO velocity analysis for selected azimuth and offset ranges by a least-squares inversion technique. The “interval approach” based on interval traveltime should be able to mitigate the impact of overlying anisotropy. The sec-ond method uses amplitude-based seismic attributes. Both methods can be implemented in a way that data are divided into a number of narrow-azimuth volumes (azimuth sector-ing). Frequently, 15, 30, or 45° azimuth bins are used. In this study, 10° azimuth sectoring was used to regularize the azimuth-offset distribution.

Both amplitude- and velocity/traveltime-based approach-es have their advantages and disadvantages (Liu et al., 2007). While the AVO response provides local, high-resolution infor-mation about anisotropy (fractures) at the top or base of the reservoirs, moveout attributes (e.g., NMO ellipse) are based on the average properties for the whole reservoir interval. We believe that when AVOAZ and azimuthal NMO analysis are combined, they may offer improved understanding of the

Figure 1. The target reservoirs consist of alternating tight dense and porous carbonate layers. The three main reservoir layers are identified as Targets I, II, and III, each being further subdivided into smaller intervals.

Downloaded 03 May 2012 to 124.81.45.4. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

July 2011 The Leading Edge 751


spatial distribution and physical properties of fractures.Figure 2 shows the AVOAZ inversion workflow used in

this study. This workflow was tested on several lines before it was applied to the entire data volume of our study area. The key steps are: superbinning and azimuth sectoring to regu-larize the azimuth-offset distribution (Figure 3); velocity fil-tering to minimize the acquisition-related linear noise (foot-print); and azimuth-offset-dependent overburden correction.

The geometry associated with seismic OBC acquisition may cause distortion in seismic amplitudes, and hence intro-duce artifacts in AVOAZ results in two ways: (1) nonuniform azimuth and offset distributions that bias anisotropy-orien-tation estimation (e.g., Holmes and Thomsen, 2002), and (2) azimuth-dependent linear noise associated with receiver and source layouts. The first problem is well known and is usually addressed by superbinning and azimuth sectoring to make the azimuth-offset distribution as regular and uniform as possible (addressed in the previous section). The second is-sue is less well understood, and can be effectively minimized using a velocity filter in the prestack azimuth domain. Veloc-ity filtering is applied in f-k space on azimuthal gathers to remove linear noise (Figure 4). Applying linear noise removal to azimuth gathers reduced amplitude scattering for AVOAZ analysis. Superbinning, azimuthal sectoring, and offset bin-ning in common azimuths enhanced the signal-to-noise ratio and regularized the offset sampling which enabled the appli-cation of velocity filtering on azimuthal gathers.

Figure 5a shows the histogram of the offset-azimuth dis-tribution before superbinning and azimuth sectoring for all traces in a crossline. As noted above, the peak number of offsets for each azimuth is clearly biased toward the source or receiver-azimuth direction. The number of offsets in the source and receiver directions is about three times more than that in the least populous azimuth sector. AVOAZ inversion was applied to each CMP, producing the orientation histo-gram shown in Figure 5b, which has two peaks close to the acquisition directions.

Similar plots are given in Figure 6a for data that have been superbinned with 50-m offset substacks. The azimuth-offset distribution no longer shows a bias toward source and receiv-er directions and is more uniform (the ratio of the maximum number of offsets to the minimum number of offsets is about two). The estimated anisotropy-orientation histogram from the superbinned data shows peaks at approximately 150° ± 20°.

Overburden correctionBecause the AVOAZ technique assumes reflection amplitude is proportional to a reservoir’s reflectivity, we must first com-pensate for any amplitude effects because of the overburden before it is applied. Complexities above target reservoirs can include regional and local structural variations, sinkholes, shallow channels, shallow anisotropy, etc. Dipping beds in general cause cos ϕ variations in apparent velocity (and there-fore traveltimes and amplitudes); shallow anisotropy (e.g., because of multiple fracture sets or multifracture layers with different orientations) may result in cos 4ϕ variations in ad-

dition to the cos 2ϕ variation. Other small-scale geological features, such as local sinkholes, channels or heterogeneities, and local topographic variations in general will cause azi-muthal amplitude scattering, resulting in poor event conti-nuity and local enhancement or dimming of seismic ampli-tudes.

For traveltime/velocity-based methods, we can obtain in-terval properties by flattening horizons below or above the target layers to obtain local measurement or by using a layer-stripping-based method. For amplitude-based methods, there is no obvious way to remove the impact of the overburden. The approach used for kinematic techniques cannot be adapt-ed, because it is unreasonable to assume a target reflector has a laterally invariant amplitude. This is a long-standing problem that has been recognized by Maultzsch et al. (2003), Luo et al. (2005, 2007), and others. None of the existing methods

Figure 2. Target-oriented AVOAZ workflow consists of optimal superbinning and azimuthal sectoring, data conditioning, linear noise reduction, and overburden correction.

Figure 3. Comparison of azimuth gathers before and after short-offset stacks and azimuthal sectoring. Trace increment is 10°.




developed to compensate for overburden effects in shear-wave anisotropy estimation are capable of removing or reducing directionally (azimuthal and offset) dependent amplitudes above target layers for AVOAZ application.

Given the difficulties of existing methods to completely remove the impact of overburden, we propose a method to remove the overburden effects for AVOAZ inversion. The method is based on the fact that any overburden effect will be imposed on some reference interval close to and above the target intervals (Figure 7). An azimuth- and offset-dependent amplitude correction transfer function can then be estimated, using for example, a Kalman filter. This transfer function is then applied as a deconvolution operator to the target events to effectively reduce the overburden effects.

The method consists of the following four key steps: (1) selection of target interval horizons; (2) selection of a refer-ence or preferably a window containing the reference horizon

above the target intervals; (3) computation of the azimuth-offset-dependent overburden-correction operator; and (4) deconvolution of the prestack data with the overburden-cor-rection operator before performing AVOAZ. The procedure can be implemented in a layer-stripping manner. This work-flow is applied in the azimuth-offset domain before AVOAZ inversion (it is also possible to implement the workflow after AVOAZ inversion).

Figure 4. Comparison of a seismic section at the azimuth of 30° before (left) and after (right) the velocity filtering, which is applied in f-k space on azimuthal gathers to remove linear noise.

Figure 5. (a) Azimuth-offset distribution for a crossline without offset binning. (b) Corresponding histogram of AVOAZ anisotropy-orientation predictions from all CMPs. Source- and receiver-line directions are indicated by red arrows.

Figure 6. (a) Azimuth-offset distribution for a crossline after 50-m (164 ft) offset superbinning. (b) Corresponding histogram of AVOAZ anisotropy-orientation predictions from all CMPs. Source- and receiver-line directions are indicated by red arrows.

Figure 7. Schematic showing approach for amplitude correction from a reference interface in order to compensate the overburden amplitude distortion. The top graph illustrates the structural variations at the shallow depth above the target interval (picked at a shallow horizon).




Figure 8. Anisotropy magnitudes before and after overburden correction. The marked area indicates a region where the anisotropy estimate is, because of the overburden, a problem that has been reduced in the lower graph.

Figure 9. (a) Anisotropy magnitude before overburden correction. Note a strong correlation between overburden geological features (e.g., the channel marked by a blue arrow) and anisotropy, as well as some periodicity. (b) Anisotropy magnitude after overburden correction. There appears to be little to no correlation between anisotropy and overburden geological features.

Figure 10. Anisotropy orientation from seismic data analysis after overburden correction overlaid on key reservoir horizon. Cyan (orange) arrow indicates direction of source (receiver) line. The rose diagrams show the fracture orientations from well analysis.

Figure 8 shows an example comparing inverted anisot-ropy magnitudes before and after the overburden correction. The striking feature in the top figure is the consistency in the anisotropy magnitude section from shallow to deep. Strong overburden anisotropy exists in the section marked with a red circle. Note that, after overburden compensation (lower fig-ure), anisotropy in the target horizon is substantially reduced in some areas, while in other areas, it is increased. In general, the method attempts to compensate for, but not completely correct for, the impact of the overburden on the azimuthal variation in target-reflection amplitude. Figure 9a shows a clear example demonstrating that, without overburden cor-rection, anisotropy determined in the reservoir interval cor-relates well with overburden karst or channels. After overbur-den correction (Figure 9b), relative anisotropy magnitudes no longer correlate with the overburden feature, indicating overburden effects have been reduced.

Results and comparison with core dataFigure 10 compares fracture orientation (color and vectors) with borehole (core and FMI) results (rose diagrams). In general, seismic anisotropy qualitatively correlates with rose diagrams at three out of five wells (A, C, and E). The anisot-ropy orientation at well H seems consistent with the second east-west fracture orientation in the rose diagram, and the main difference is that seismic “fracture” orientation is dom-inated by north of west and south of east orientations, while borehole data (rose diagrams) give mainly NNE and SSW orientations. Also note that there is no observable anisotropy around well I which is consistent with the results from the analysis of 3D VSP data in this well (not shown here).




ConclusionsWe have proposed a workflow and method to mitigate over-burden effects in order to use AVOAZ analysis to extract geologically meaningful fracture distributions at reservoir levels. We conclude that (1) data quality is crucial to the success of this technology, and a carefully designed process-ing flow is needed to effectively improve seismic images and reduce noise, while preserving relative azimuth information for anisotropy-attribute inversion; (2) seismic-anisotropy at-tributes extracted at reservoirs are strongly affected by the acquisition footprint (because of source-receiver distribution, linear noise associated with source and receiver spacing), which needs to be minimized; and (3) shallow geological fea-tures (e.g., karst, channels, and other small scale variations) can significantly overshadow the extracted attributes at the target level. We developed a method to reduce these overbur-den effects while preserving the relative information at the reservoir levels. Finally, (4) seismic “fracture” attributes need to be integrated with other geological and geophysical data (e.g., borehole data) in order to be able to extract meaningful results.

ReferencesHolmes, G. and L. Thomsen, 2002, Seismic fracture detection at a

Middle East offshore carbonate field: SPE paper 78507.Liu, E., M. Chapman, I. Varela, X. Li, J. H. Queen, and H. Lynn,

2007, Velocity and attenuation anisotropy: implication of seismic fracture characterizations: The Leading Edge, 26, no. 9, 1170–1174, doi:10.1190/1.2780788.

Luo, M., I. Takahashi, M. Takanashi, and Y. Tamura, 2005, Improved fracture network mapping through reducing overburden influence: The Leading Edge, 24, no. 11, 1094–1098, doi:10.1190/1.2135091.

Luo, M., M. Takanashi, K. Nakayama, and T. Ezaka, 2007, Physical modeling of overburden effects: Geophysics, 72, no. 4, T37–T45, doi:10.1190/1.2735925.

Maultzsch, S., S. Horne, S. Archer, and H. Burkhardt, 2003, Effects of an anisotropic overburden on azimuthal amplitude analysis in horizontally transverse isotropic media: Geophysical Prospecting, 51, no. 1, 61–74, doi:10.1046/j.1365-2478.2003.00354.x.

Reilly, J. M., A. P. Shatilo, and Z. J. Shevchek, 2010, The case for separate sensor processing: Meeting the imaging challenges in a producing carbonate field in the Middle East: The Leading Edge, 29, no. 10, 1240–1249, doi:10.1190/1.3496914.

Acknowledgments: We thank all shareholders for support. The re-sults achieved would not have been possible without the continued advice, support and guidance from ADNOC, INPEX-JODCO, ADMA-OPCO, EMAD, and ZADCO staff and management. We thank guest editors Reinaldo Michelena and Heloise Lynn for constructive comments.

Corresponding author: [email protected]


Azimutal Avo

Documents

seismic data

seismic fracture attributes

seismic fracture prediction

fracture clusters

i t i s

fractureanisotropy attributes

acquisition footprint

seismic amplitude