Seismic attributes and acoustic impedance inversion in interpretation of

complex hydrocarbon reservoirs

Mohammed Farfour, Wang Jung Yoon ⁎, Jinmo Kim

(Geophysical Prospecting Lab, Energy and Resources Engineering Department, Chonnam University, Yong-Bon Dong, Gwangju 500-757, South Korea)



Thin-bed reservoirs,Seismic attribute,Spectral decomposition,Acoustic impedance

 لایه های نازک مخزنی ,خواص و ویژگیهای لرزه ای,تجزیه طیفی,امپدانس آکوستیک,وارونه و معکوس


Seismic attributes along with seismic inversion are used to study a challenging reservoir facies from Boonsville field in north central Texas. Two thin reservoirs, Upper Caddo and Lower Caddo, separated by a nonproductive thin limestone formation are investigated. Seismic was not an unambiguous indicator of the Caddo reservoir.The Caddo event exhibits doublet character reflections when productive facies are present. However, this character proved to be a false diagnostic of the Caddo reservoir. Wells drilled on interpreted doublet reflections turned out to dry holes penetrating two thin limestone units instead.First, some modern poststack processing techniques are used to remove incoherent noise and improve the data resolution and interpretability. Seismic instantaneous attributes (i.e., amplitude, phase, and frequency) are then calculated to infer seismic expressions of the reservoir and the limestone formations. Next, spectral decomposition is performed to study the frequency responses of the different lithotypes. Interestingly, the attribute images and frequency maps could distinguish between the different investigated units. However, some limestone units showup similar to the Lower Caddo formation in the frequency images.Model-based acoustic impedance inversion is carried out on seismic lines spanning both zones. Acoustic impedance sections reveal that producingwells and dry wells penetrate different formations. The cosine of instantaneous phase attribute along with impedance sections supports the above interpretations and displays more clearlywhere the Caddo reservoir is traversed and where it is not. This study demonstrates that integration of modern posts tack processing and novel seismic attributes can lead to extract more key information about challenging reservoirs with less ambiguity.

 Geological background:

Boonsville field is located in the Fort Worth basin of north central Texas . The reservoir system we study here is the Bend Conglomerate, a productive series of gas reservoirs composed of Middle Pennsylvanian fluvio-deltaic clastics about 900 to 1300 ft (275 to 400 m) thick in our project area, with the base of the interval being a little less than 5000 ft (1525 m) deep. We focus particularly on the Caddo interval, which is one of the most productive reservoirs in the area. The Caddo contains two discrete sandstone bodies: one within a Lower Caddo between the maximum flooding surface MFS80 and Caddo limestone, and one in an Upper Caddo between the other maximum flooding surface MFS90 and Caddo limestone. The reservoir zones are typically characterized by noticeably low Spontaneous Potential (SP) and Gamma Ray (GR) as a result of low content in radioactive and clay minerals. High resistivity (Res) values caused by hydrocarbons are commonly seen at these zones . Most of the Caddo production in our study area is oil, however, significant gas reserves have been found in some Upper Caddowells that are located structurally downdip relative to the Lower Caddo oil accumulations, indicating that the two units are physically separate reservoir compartments (Hardage et al., 1996b). Hardage et al. (1996a,b) analyzed the reservoir distribution and depositional subfacies using a few typical wells from this study area, buttheir model is not sufficiently detailed to be useful for facies analysis of this thin Caddo sequence.Themain delta system was divided into four delta subfacies: 1) distal channel in the east margin; 2) proximal delta front in the east; 3) distal delta extending to south and northeast; and 4) inter delta in the south and southeast. The sandstone reservoirs are widely distributed in the delta front subfacies.

Fig. 1. Workflow followed to prepare the data for seismic attribute and horizon tracking.



Our primary seismic imaging objective is to delineate the Lower and Upper Caddo reservoirs boundaries and to distinguish sand-fill from limestone-fill. To enhance the resolution and remove incoherent noises from the data, dip-steered coherency filtering workflows were employed (Fig. 3). Dip-steering-based filtering is the process of following the local dip and azimuth to find the input segments for the filtering process (Davies et al., 2004). Accordingly, a dip-steering cube is generated that calculates local dip and azimuth at every sample position within the seismic. The smoothed steering cube is subsequently integrated to calculate dip-steered median filter and dip-steered average filter. These dip-steered filters can help to significantly improve data interpretability and smooth seismic data along reflection dips while preserving faults and other sharp discontinuities (e.g., Barnes et al., 2012). Furthermore, for tracking our target horizonwith less uncertainty, Horizon Cube, a very dense tracking approach, is invoked (De Grout et al., 2010). The same smoothed steering cube is used to generate a set of auto-tracked horizons that are typically separated by one sample apart on average. After that, the user can isolate any horizon at any point. As a result, interpretation of densely tracked horizons becomes much easier and more reliable with filtered data, and seismic attributes appear much cleaner. Fig. 4 displays the tracked horizon before and after poststack processing. It is important to note that both reservoirs and the thin bed limestone that separates them are found within 30 ms starting from the tracked horizon. We perform our analysis over the time interval corresponding with the targeted reservoirs. After that, we run spectral decomposition using Continuous Wavelet Transform (CWT). In practice, the CWT approach involves the following steps: 1) decompose the seismogram intowavelet components as a function of the scale σ and the translation shift τ; 2) multiply the complex spectrum of each wavelet used in the basis function by its CWT coefficient, and sum the result to generate instantaneous frequency gathers; 3) and these gathers are then sorted to  produce constant frequency cubes, time slices, horizon slices, or vertical sections (Chopra and Marfurt, 2007). In our case, the Mexican hat wavelet was selected among other available wavelets. Then, CWT images were calculated by setting the attribute at an initial frequency component which will change (increase or decrease) by predefined increments. Each frequency component is expected to enable the interpretation of subtle details of the stratigraphic framework of the reservoirs. To overcome ambiguities arising from spectral decomposition, some seismic lines crossing some key producing wells and non producing wells are converted to acoustic impedance sections. Results from inversion and the cosine of instantaneous phase attribute applied to relative impedance do support the interpretation derived from spectral decomposition images.We show that integration of innovative techniques of poststack processing,  advanced autotracking, and attribute analysis can lead to promising results for studying subsurface formations.


 Fig. 2. The survey spectrum shows the zone of interest frequency content. Time interval is from 600 to 1200 ms. 

Fig. 3. Target horizon with and without filtering.



 Results and discussion:

Examination of the Boonsville 3-D seismic data showed that the seismic reflection associated with the Caddo sequence boundary underwent a significant amplitude reduction along the trend where the well-log-based map shows that the Upper Caddo reservoir sandstone existed. Fig. 5 displays the spectrum of the zone of interest. The tuning analysis shows that the seismic cannot resolve formations thinner than a temporal thickness of 8 ms,which is equivalent to 120 ft. According to engineering data, the Caddo reservoirs do not exceed 100 ft in thickness. Consequently, seismic analysis was not an unambiguous tool in characterizing this challenging reservoir. Moreover, it is noticed that the Caddo reflection event develops a doublet characterwhen productive facies are present. However, these doublet reflections found a misleading character at the northwest area. Figs. 6 and 7 showtwo seismic sections passing by two different wells. The first well (I.G.Y.13) is a producing well and the other one (B.Y.18D) is a dry well. The seismic characters at the producing and drywells look similar, although the productive Caddo facies was not penetrated at B.Y.18D. Results from previous work by Hardage et al. (1996a) pointed out that some areas of Caddo reservoirs can be defined using seismic attributes. While an averaged instantaneous frequency map could predict some parts of the Lower Caddo and show results that are somewhat consistent with well-based maps, averaged instantaneous amplitude failed to predict unambiguously the Upper Caddo reservoir facies distribution and presented the limestone beds as productive sandstone.We expected that spectrally decomposing the seismic data and running a careful analysis may help solve this problem. Therefore, we carried out a wavelet-based spectral decomposition. For this decomposition all available wavelets have been examined. The Mexican hat wavelet is found to be the best in terms of lateral frequency and vertical time resolutions compared to Morlet and Gaussian wavelets. Several frequencies have been calculated for the carefully tracked Caddo horizon. At 50 Hz the reservoir horizon frequency amplitude shows trends somewhat similar to what was observed in the instantaneous frequency map , notice that the 50-Hz frequency image could map not only the Lower Caddo but also the Upper Caddo as it is shown in the northeastern part. However, it is observed that two reservoirs appear with distinct characters. It can be readily seen that the producing wells penetrate different zones suggesting that the two reservoirs are physically independent units. That is, AC1, AC2, AC3, AB3, and C.B21.1 (in yellowmark) are producing fromone zone (Upper Caddo),whereas I.G.Y.3, I.G.Y.31, I.G.Y.13, I.G.Y.21, I.G.Y.19, L.O.F1, L.O.F2, L.O.F3, and L.O.F.5 (in black) are producing from another productive zone (Lower Caddo). The reason that the two reservoirs are both apparent on the frequency image is due to the fact that the reservoirs are thin (below tuning) and so close vertically that their images can be seen on one horizon. According towell data, both reservoir intervals are presentwit hin a time window not exceeding 30 ms. In such situations, geologic features, or at least most of them, can be seen on more than one closely spaced horizon. It is up to the interpreter to choose the interval where the feature analyzed is best resolved (Roksandic, 1995).



 Fig. 4. A seismic crossline traversing the B.Y.18D well. Note that this dry well penetrates two thin nonproductive limestone units.


Fig. 5. Average instantaneous seismic frequency calculatedwithin the Lower Caddo sequence. The map is found to be coincidingwith the Lower Caddo net reservoir sandstonemap.Wells

in black are producing fromLower Caddo.High frequency in Lower Caddo zone is found coincidingwith net reservoir (red circle). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)



Fig. 6. 50 Hz horizon shows Lower and Upper Caddo reservoir distributions. All Lower Caddo producing wells (in black) lie in low amplitude, while Upper Caddo producing wells (in yellow)

fall in high amplitude. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 

Furthermore, we remarkably noticed that all producing wells from Lower Caddo reservoir are found characterized by low frequency amplitude trending northeastward,while wells producing from Upper Caddo are found to be associated with high spectral amplitude.Experience showed that geological formations are characterized by their specific frequency behaviors. This is due partly to their unique rock and fluid properties in their surrounding environments (Tai et al., 2009; Chen et al., 2008).



Fig. 7. Workflow of performed inversion. 

It is important also to note that there exist other areas,where neither Caddo's lower nor upper reservoir is penetrated, that show relatively similar frequency responses to those discussed above. That is, a low amplitude spectrum anomaly passing through the wells B.Y2, B.Y3, and B.Y18D is found similar to that of the Lower Caddo. Thus, the anomalously low amplitude spectrum is not a sufficient diagnostic of hydrocarbon-charged Caddo formation. This situation is a typical example of the non uniqueness of frequency responses, which implies that more than one stratigraphic condition can create the same spectral response. In order to solve this problem and provide more support to our interpretation, seismic model-based inversion is carried out using Hampson-Russell commercial software. The objective is to derive acoustic impedance sections from the integration of both well logs and seismic data.  depicts the workflow of the inversion process. The sections traverse both producing and non producing zones. Only 4 wells have been integrated in the inversion process. Of the four wells, one well did not contain a sonic log; thus, the log was generated with the help of the log data from the other wells. The four wells have been correlated and tied to the seismic traces at well location to produce acoustic impedance sections. An excellent match between syntheticand real data has been reached with minimum misfit error estimatesAs expected, the acoustic inversion sections provided more key information than seismic data.  interestingly, the impedance sections demonstrate that producing wells do pass through the Upper Caddo reservoir characterized by their high impedance, while the dry well (B.Y.18D) appears to cross other formations with relatively lower impedance; they are the two thin-bed limestone units. The incised valley fill reservoirs in our area are known for their relatively higher density and velocity compared to their surrounding rocks. The cosine of instantaneous phase of the relative acoustic impedance was then calculated and blended with impedance data. This attribute is basically used in seismic interpretation to follow seismic event continuity and detect reflection terminations.We have applied the attribute here on the relative impedance, instead, to follow the impedance character associated with the Caddo reservoirs. display two random lines passing through the wells investigated above. It is clearly seen that the impedance character associated with the Caddo reservoir does not reach the well B.Y.18D.

Fig. 8. Real vs. synthetic data and misfit error derived from the inversion. 



Fig. 9. Impedance section crossing the Lower Caddo producing well I.G.Y.13.

On the contrary, at wells I.G.Y.31 and I.G.Y.21, the high Caddo reservoir impedance character is passed by both wells and demonstrates a good continuity. The attribute with the help of acoustic impedance does not only show the continuity of the reservoir, but it reveals unambiguously where the reservoir is present and where it is not. Therefore, this does not only support the spectral decomposition results, but it removes remaining ambiguity in achieving our objective.


Seismic attributes are used to map thin Caddo reservoirs and differentiate their distribution fromthe limestone units. The incorporation of cleaner and interpretive seismic attributes resulted in successfully overcoming several difficulties encountered throughout the study of these challenging reservoirs.While instantaneous attributes (amplitude and frequency) assisted in defining some expressions of the Caddo target, spectral decomposition could reveal stratigraphic features of the different formations and discriminate them from each other. In order to provide more support to the above results and remove raised ambiguities, seismic inversion has been performed to integrate information from wells with seismic data and derive acoustic impedance sections. Results from the inversion do confirm our interpretation and were found to be consistent with spectral decomposition achievements. Cosine of instantaneous phase combined with impedance revealed further information and demonstrated where the reservoir is penetrated and where it does not exist. The study demonstrates how integration of modern poststack processing (dip-steered filtering and tracking) and novel approaches (spectral decomposition, model-based seismic inversion) can help exploit different available data and information in solving problems encountered in subsurface imaging and characterization.

Fig. 10. Impedance section crossing two wells I.G.Y.21 and 31 producing from the Lower Caddo



(spectral decomposition, model-based seismic inversion) can help exploit different available data and information in solving problems encountered in subsurface imaging and characterization. Acknowledgments Wewould like to thank dGB Earth sciences Company and the people from for providing the software (Opendtect) and also Hampson-Russell Software Company for providing the software (Strata). Our thanks go as well to the Texas Bureau of Economic Geology for providing the data set and to Seismic Micro-Technology (SMT) for providing the software (Kingdom). Finally,we express our gratitude to the Korean Government for their support under the BK21 Plus program.

Fig. 11. Impedance section passing through the B.Y.18D. Two thin limestone units have been penetrated instead of the incised valley reservoir. 


Fig. 12. Cosine of instantaneous phase of the relative impedance blended with absolute impedance section revealing that the Caddo reservoir was not encountered by the B.Y.18D.

Fig. 13. Cosine of instantaneous phase of impedance showing the continuity of the reservoir penetrated by the two wells. 




Barnes, A.E., Rahim, M.F.A., Ghazali, A.R., 2012. Workflows to improve the resolution of

post-stack seismic data and attributes from the Malay Basin. Petroleum Geoscience

Conference & Exhibition.

Castagna, J.P., Sun, S., Siegfried, R.W., 2003. Instantaneous spectral analysis: detection of

low‐frequency shadows associated with hydrocarbons. Lead. Edge 22, 120–127.

Chen, G., Matteucci, G., Fahmi, B., Finn, Ch., 2008. Spectral‐decomposition response to reservoir

fluids from a deepwater West Africa reservoir. Geophysics 73, C23–C30.

Chopra, S., Marfurt, K.J., 2007. Seismic Attributes for Prospect Identification and Reservoir

Characterization. Society of Exploration Geophysicists, Tulsa, OK (456 pp.).

Davies, R.J., Cartwright, J.A., Stewart, S.A., Lappin, M., Underhill, J.R. (Eds.), 2004. 3D Seismic

Technology. Application to the Exploration of Sedimentary Basins 29. Geological

Society, London, Memoirs.

De Grout, P., Huck, A., de Bruin, G., Hemstra, N., Bedford, J., 2010. The horizon cube: a step

change in seismic interpretation. Lead. Edge 29, 1048–1055.

Farfour, M., Yoon, W.J., 2013. Ultra-thin bed reservoir interpretation using seismic attributes.

Arab. J. Sci. Eng. 39 (1), 379–386.

Hardage, B.A., Carr, D.L., Lancaster, D.E., Simmons, J.L., Hamilton, D.S., Elphick, R.Y., Oliver,

K.L., Johns, R.A., 1996a. 3D seismic imaging and seismic attribute analysis of

genetic sequences deposited in low accommodation conditions. Geophysics 61,


Hardage, B.A., Simmons, J.L., Lancaster, D.E., Elphick, R.Y., Edson, R.D., Carr, D.L., 1996b.

Boonsville 3-D Seismic Data Set: The University of Texas at Austin. Bureau of Economic

Geology (32 pp.).

Latimer, R.B., Davidson, R., van Riel, P., 2000. An interpreter's guide to understanding and

working with seismic‐derived acoustic impedance data. Lead. Edge 19, 242–256.

Lavergne, M., Willm, C., 1977. Inversion of seismograms and pseudo-velocity logs.

Geophys. Prospect. 25, 232–250.

Lindseth, R.O., 1976. Seislog process uses seismic reflection traces. Oil Gas J. 74 (43),


Partyka, G., Gridley, J., Lopez, J.A., 1999. Interpretational applications of spectral decomposition

in reservoir characterization. Lead. Edge 18, 353–360.

Pendrel, J., 2006. Seismic Inversion — Still the Best Tool for Reservoir Characterization:

CSEG Recorder vol. 31, no. 1.

Pennington, W.D., Acevedo, H., Green, A., Haataja, J., Len, S.,Minaeva, A., Xie, D., 2002. Calibration

of seismic attributes for reservoir characterization. Final Report MTU,

Houghton (185 pp.).

Puryear, Ch.I., Castagna, J.P., 2008. Layer-thickness determination and stratigraphic interpretation

 using spectral inversion: theory and application. Geophysics 73 (2), R37–R48.