The coherence attribute is an edge detection method that is widely used for interpreting faults on 3-D seismic time slices. The traditional coherence attribute is calculated on migrated volumes using traces from all available azimuths. It has recently been shown that calculating coherence along specific azimuths can enhance the detection of faults running perpendicular to those azimuths. In this study, we applied azimuthal coherence attribute analysis on a 3-D seismic data set from a gas field in Central Saudi Arabia. We generated four migrated 3-D data volumes sorted by azimuth in addition to a conventional full-azimuth volume. We then calculated the coherence attribute for all volumes and compared each azimuthal coherence volume to the conventional full-azimuth coherence volume. The azimuthal coherence results exhibited an improved definition for faults whose strikes are perpendicular to the sorting azimuth. More specifically, systems of NW-trending discontinuities were imaged more clearly in the NE-SW oriented coherence volume than it was in the full-azimuth coherence volume. The reason for this enhancement is the fact that seismic waves tend to avoid passing through the fault when they propagate parallel to the fault strike therefore missing the effects of the fault while they must pass through the fault when propagating perpendicular to the fault strike which results in better illumination of the fault.

One of the most important steps in seismic data interpretation is the ability to map structural faults reliably. Seismic attributes, such as coherence, are frequently used to help delineate these faults. It is common practice to interpret faults using constant or horizon time/depth slices extracted from 3-D coherence volumes; coherence being defined as a measure of similarity between reflector waveforms. Different concepts of coherence are described by researchers such as Drecun and Lucas (1985) and Claerbout (1990). One of the earliest documented examples of the use of coherence for fault mapping was the application of a three trace cross-correlation coherence algorithm to a large 3-D seismic survey acquired over South Marsh Island, Louisiana, USA (Bahorich and Farmer, 1995). In a 3-D coherence time slice the coherence along a fault plane decreases because the similarity between the traces across the fault is low, i.e., incoherent. As a result, coherence can reveal significant structural and stratigraphic boundaries because these features generate sharp discontinuities that are reflected on the seismic traces.

Gersztenkorn et al. (1999) showed an excellent example of the application of coherence technology using a 3-D seismic survey located in Trinidad. In this survey, they demonstrated that using full-azimuth seismic amplitude time slices over a complex fault region only detected faults that ran perpendicular to the seismic inline direction whilst all other faults, regardless of orientation, were generally better imaged on a coherence time slice. Chopra (2002) introduced a new methodology where the pre-stack data was sorted into different azimuthal ranges followed by stacking and migration of these azimuthal sector volumes, after which the coherence attributes were calculated. Al-Dossary et al. (2004) further refined this sectoring approach by additionally sorting the pre-stack data by offset range prior to stacking, migration, and the generation of the coherence volumes.

There are a number of examples where seismic attributes have been used in the interpretation of structural and stratigraphic features in Saudi Arabia. Bremkamp et al. (2004) used coherence and spectral decomposition attributes, in addition to the traditional geometric horizon attributes of dip and azimuth, to map subtle faults and fractures in the Upper Permian Unayzah sandstone of Central Saudi Arabia. Al-Zahrani and Neves (2008) used seismic attributes in the northern part of Central Saudi Arabia to delineate the aeolian sand deposits and the fluvial channels of the Unayzah Formation. They used coherence and curvature attributes to map a fault and a fluvial channel network at the top of the Unayzah Formation.

The work by Al-Hawas et al. (2003) characterized fractures in the Wudayhi Field using P-wave azimuthal velocity and AVOA analysis. They documented a small azimuthal variation in P-wave velocity and a major variation in AVOA at the reservoir level, which indicated the existence of vertically aligned fractures. They concluded that the predictable orientation of maximum principal horizontal stress (PHS) from azimuthal velocity and AVOA analyses is E-W.

The detection of faulting and fault orientation is a challenge for all explorationists because conventional full-azimuth stacking tends to smear fault effects due to the mixing of dip and strike azimuths that has different illumination degrees of the fault. The objective of this study is to use azimuthal coherence attribute analysis for mapping structural features such as faults, particularly those that are parallel to travel path azimuths. This was achieved by implementing the following 3 steps:

  1. 3-D seismic data from a field in Central Saudi Arabia was sorted into N-S, E-W, NE-SW, and SE-NW azimuths (Figure 1).

  2. Coherence was calculated on these sub-volumes as well as on the full-azimuth volume.

  3. The four azimuthal-limited volumes were compared to the full-azimuth volume.

The azimuthal coherence technique was applied to a 3-D seismic data set from a gas field in Central Saudi Arabia. This field has been studied by several researchers who focused on mapping the existence and distribution of fractures using 3-D surface seismic and vertical seismic profiling (VSP) data (Al-Hawas et al., 2003; Neves et al., 2003; Nebrija et al., 2003; Owusu and Nebrija, 2007). The field is an asymmetric structural anticline made up of Paleozoic–Cenozoic era sequences with a gentle southern dip of several degrees. Well tests showed that productivity can be enhanced by identifying the distribution and orientation of fractures in the reservoir, particularly in the lower part of the Unayzah B reservoir (Figure 2) (Al-Hawas et al., 2003).

The Upper Carboniferous–Lower Permian Unayzah Formation in the field is composed of three main units: Unayzah-A (eolian), Unayzah-B (fluvial and lacustrine) and Unayzah-C (glaciofluvial/eolian) units. The Unayzah sedimentology and stratigraphy were described in detail by Senalp and Duaiji (2001).

Core and borehole image logs data within the Unayzah-A reveal no open fractures, compared to the Unayzah-B, due to changes in lithology (Figure 2). Early soft sediment deformation, faulting, granulation seams and slump structures are observed in the Unayzah-A. The Unayzah-B has mainly two types of fractures: stylolite related and non-stylolite related. These fractures are mostly open with no mineralization, although some mineralization has been observed (e.g., Al-Hawas et al., 2003).

The Rose diagrams from two wells in the field, at the Unayzah-B level, show two dominant vertical sets of open fractures oriented ENE-WSW and NW-SE (Figure 3). The horizontal in-situ stress has been derived from borehole breakouts and drilling-induced tensile fractures (Ameen and Hailwood, 2008). The closed fractures are oriented NNE and the soft-sediment faults are oriented NNW within Unayzah-A (Al-Hawas et al., 2003).

The Unayzah Formation is cut by an EW-trending fault zone that is consistent with the regional tectonic trend observed on 3-D surface seismic, VSP and borehole breakout data (Al-Hawas et al., 2003; Neves et al., 2003; Nebrija et al., 2003; Owusu and Nebrija, 2007).

The seismic data used in this study was obtained from a conventional 3-D Vibroseis survey acquired in the Central Province of Saudi Arabia in 1999. Saudi Aramco acquired the 480-fold 3-D survey (Table 1), which covered approximately 300 sq km, to improve the imaging of the Unayzah Formation. The general seismic processing sequence used in the analysis is shown in Figure 4. The main processing steps included: geometry, near-surface modeling, noise suppression, surface-consistent deconvolution and amplitude corrections, residual static corrections, velocity analysis, azimuthal sorting and stacking, and post stack migration.

Common problems with seismic data were coherent and random noise. Significant efforts were made during processing to suppress the noise as much as possible to avoid any artificial edges or boundaries in migrated volume. A linear adaptive noise attenuation module was found to be the most effective amongst several methods tested. Next, surface-consistent scaling and deconvolution were applied to avoid amplitude and phase distortions. Prior to migrating the data, a velocity-driven multiple suppression methodology was applied. This method suppressed both surface-related and interbed multiples that exhibited a different stacking velocity from those of primaries.

An azimuth-sorting module was used to sort the data into different azimuths. This module calculated azimuth and offset values for 3-D data from the X and Y trace coordinates and these calculated parameters were then stored in the trace headers. Four limited-azimuth range volumes were selected from the full-azimuth 3-D pre-stack seismic data. The azimuth directions selected were N-S (±45°), E-W (±45°), NE-SW (±45°), and NW-SE (±45°). An important parameter was the selection of the range of azimuthal data. For example, for selecting data in the NE-SW direction, it was necessary to specify two azimuth ranges from 0° to 90° and 180° to 270°. After each subset of azimuth-limited data was sorted into individual volumes, the volumes were stacked.

The coherence attribute was then calculated from the four azimuth-limited volumes and the full-azimuth volume after applying post-stack time migration using Kirchhoff time migration. A Sobel edge detection algorithm had been used to map subsurface geological discontinuities such as faults, channels, and fractures (Gonzalez and Woods, 1992).

Full-Azimuth Seismic Data

This section focuses on data examples taken from the full-azimuth volume. Time slices from the top and bottom of the Unayzah reservoir were selected for comparison. The seismic data quality is good. The anticlinal structure can be clearly seen from both the in-line and cross-line directions through the full-azimuth volume (Figure 5). Figure 6 is a zoomed view around the reservoir level. The two lines taken from the middle of the field are displayed in green (in-line) and blue (cross-line) as shown in Figure 7.

Some differences were observed when comparing time slices from the full-azimuth amplitude volume and from the full-azimuth coherence volume. While the time slices from the original amplitude volume did not clearly delineate a major fault boundary, the major fault could be clearly seen using the coherence time slices, and these slices also revealed the presence of additional subtle faulting (Figure 7). A system of faults trending NE-SW was difficult to see in the amplitude time slice but was apparent in the coherence time slice. Some subtle discontinuities seen in the coherence time slice at the top of the Unayzah reservoir exhibited a narrow coherence lineation as indicated by the red arrows in Figure 8. At the base of the Unayzah reservoir, these lineations (also indicated by red arrows), were broader and smeared, possibly due to noise in the data.

Azimuth-limited Seismic Data

Four azimuth-limited seismic volumes were generated (NE-SW, NW-SE, E-W, and N-S) with nominal fold of 253, 262, 115, and 405, respectively. The fold is not the same in all azimuths because the acquisition geometry emphasizes N-S azimuths. An examination of an in-line (Figure 9) and cross-line (Figure 10) from the four azimuth-limited seismic volumes showed image quality varying with azimuth due to the presence of generally E-W trending faults. Comparison of coherence time slices at the top (Figure 11) and base (Figure 12) of the Unayzah reservoir taken from the azimuth-limited seismic volumes showed that fault delineation was generally enhanced on azimuth-limited coherence volumes that were perpendicular to the fault strike.

Comparison of the Full-azimuth and Azimuth-limited Coherence Volumes

Coherence slices generated from the full-azimuth volume did not reveal the low-coherence features (i.e. subtle discontinuities) associated with deformation of the reflectors. In contrast, coherence generated from an azimuth-limited volume generally enhanced low-coherence features in directions perpendicular to the binning azimuth. While the overall signal-to-noise ratio (SNR) and fold of each azimuth-limited volume was less than that of the full-azimuth volume, the fault imaging was enhanced after coherence was run compared to the full-azimuth volume.

The NE-SW azimuth-limited coherence slice at the top of the Unayzah reservoir showed that the major ENE-trending fault was almost smeared because the fault ran parallel to the binning azimuth direction, as indicated by the red arrow (Figure 13a). The E-W azimuth-limited coherence slice improved the imaging of the N-S trending faults. Few subtle discontinuities were apparent in the northern part of the slice as indicated by the green arrow (Figure 13c), which were not obvious in the full-azimuth slice (Figure 13e).

Although the overall seismic data quality of the E-W azimuth-limited volume was better than the N-S azimuth-limited volume, more discontinuous features were observed in the E-W azimuth-limited coherence slice. At the bottom of the Unayzah reservoir, two distinct zones, indicated by the green circles, were visible in the E-W azimuth-limited coherence slice that might indicate the existence of N-S trending fracture swarms (Figure 14c). A significant NW-SE trending feature from the coherence time slice through the NE-SW azimuth-limited volume was also observed (Figure 14a). This azimuth-limited coherence slice showed that the northwest-southeast trending subtle discontinuity indicated by the yellow arrows intersected the major ENE fault trend.

This study of fault mapping using azimuthal coherence yielded three significant findings:

  1. It demonstrated the superiority of using coherence time slices over amplitude time slices on full-azimuth seismic data.

  2. Coherence time slices taken from the top and bottom of the Unayzah reservoir from four azimuth-limited volumes showed various subtle faults with different orientations.

  3. The ability to interpret and map faulting is enhanced when using azimuth-limited volumes rather than the full-azimuth volume.

The main objective of this study was to map and identify reservoir zones where gas productivity could be influenced by the presence of subtle faults, which would enhance porosity and permeability. Using azimuthal coherence provided an opportunity to compare coherence time slices from different ranges of azimuthal data to the full-azimuth coherence volume to look for areas with better focusing and better delineation of both larger discontinuities and smaller more subtle features. Each azimuth-limited volume revealed faults which were perpendicular to the sorting azimuth. Although the SNR of each azimuth-limited volume and fold were significantly less than that of the full-azimuth volume, the lateral resolution was enhanced compared to the full-azimuth volume. The results of this study will play an essential role in placing future development wells in this field.

We are grateful to King Fahd University of Petroleum and Minerals for supporting this study. We would like to express our sincere gratitude to Saleh Al-Dossary and Mustafa Hariri for their helpful comments. This work would not have been possible without the support of Saudi Aramco, especially the Geophysical Data Processing Division of the Exploration Operations Department, permitting us to process, interpret and publish the data used for this study. The useful suggestions by an anonymous reviewer and the design by GeoArabia Graphic Designer Nestor ‘Nino’ Buhay IV, are greatly appreciated.


Faisal Al-Qahtani earned his BSc (2003) and MSc (2009) degrees in Geophysics from King Fahd University of Petroleum and Minerals (KFUPM), Saudi Arabia. In 2004 Faisal joined Saudi Aramco, where he has worked in seismic data acquisition, processing, and interpretation. His research interest is in processing and interpreting seismic data. He is a member of the SEG, EAGE and the Dhahran Geoscience Society.

Abdullatif Al-Shuhail is an Associate Professor of Geophysics at King Fahd University of Petroleum and Minerals (KFUPM), Saudi Arabia. He received a BSc degree in Geophysics from KFUPM in 1987, and worked as a Graduate Assistant at the Department of Earth Sciences between 1988 and 1990. Abdullatif obtained an MSc in Geophysics in 1993, and a PhD in Geophysics in 1998, from Texas A&M University, College Station. His fields of interest include seismic characterization of fractured reservoirs and near-surface effects on petroleum seismic data. Abdullatif is a member of the SEG, EAGE, and Dhahran Geoscience Society.