SBGf Sociedade Brasileira de Geofisica September 2003

Spectral Decomposition Reveals Geological Hidden Features in the Amplitude Maps from a Deep Water Reservoir in the Campos Basin
Paulo Johann and Gilberto Ragagnin PETROBRAS S/A Mrcio Spnola Landmark Graphics Corporation This work attempts to present the spectral decomposition as a qualitative and quantitative tool to indicate geological discontinuities and images of the thickness variability, with a focus on the statistical aspects of the frequency slices.
Spectral decomposition has been frequently used by the geophysical community in seismic processing workflow, such as; spectral analysis, filtering, signal analysis, etc. Only recently has spectral decomposition been used for the interpretation of seismic data (Partyka et al., 1998; 1999; Bahorich et al., 2001). This work attempts to present the spectral decomposition as a qualitative and quantitative tool to indicate geological discontinuities and images of the thickness variability, with a focus on the statistical aspects of the frequency slices. The technology used in this work was originally developed by researches from Amoco (Partyka and Gridley, 1997), and later by BP and Apache in partnership with Halliburton/Landmark. In Petrobrass research project New Technologies Applied to Reservoir Characterization of Thin Turbidites Reservoirs, PRAVAP 19 - Advanced Oil Recovery Program, there is a special interest in the use of this technology in order to obtain an estimation of the thickness especially in the presence of stratigraphic pinch-outs that may be causing tuning effect in some reservoirs. These geological situations represent important oil volumes in the main reservoirs of the company. This method was recently developed to represent the tuning effect as an important characteristic of spectrum to be revealed and not as an effect to be removed by traditional procedures of seismic processing. From the interpreters perspective the richness of this method is the capacity to reveal seismic stratigraphy and seismic structural features that cannot be seen in full-band width dominant frequency (time or depth domain). Today most seismic interpreters usually work with amplitude anomalies based on dominant frequency. With the method presented here we emphasize the added value of working with the discrete components of seismic bandwidth. Each frequency component is available to help to understand and interpret subtle seismic structural behavior and new details of the stratigraphic framework of the oil field.

Abstract
Results and applications of the spectral decomposition based on the Discrete Fourier Transform (Bracewell, 1986 e Gridley e Partyka, 1997) are presented in this article. An emphasis on the statistics and quantitative aspects obtained from frequency slices of the 3D tuning cube are addressed. Fourier transform is characterized by a linear transform, frequently used to solve mathematical problems representing natural phenomena in a different domain where the solution couldnt be found or was not completely resolved. This work attempts to show an application of spectral decomposition in deep water Campos Basin project, in the presence of channel-lobes turbidite play. Faults are also present and are potential boundaries for the hydrocarbon flow.

Spectral Decomposition Methodology

The method of spectral decomposition through Discrete Fourier Transform (DFT) presented in this work is based on a phase independent, qualitative and quantitative approach of spectral decomposition (Partyka, 1999). Seismic reflections, particularly from thin bed reservoirs, have characteristic expressions in the frequency domain.

Introduction
Many methods have been applied to seismic data in the reservoir characterization domain in order to produce images that can reduce the uncertainties associated to the geologic interpretation process, litho-stratigraphic interpretation and fluid contents.

Landmark Technical Review 75

These features on a map view of the amplitude spectrum were obtained from the tuning volume representing the time tuning thickness of the geological beds. The workflow consisted of choosing an interval of interest in time domain (3D slab containing sequence of interest) that was flattened and then through DTF, transformed into frequency domain (Figure 1a - c). In order to compare different frequencies, a spectral balancing of amplitudes was performed (Figure 1 d). Animating through the spectral decomposition images allows interpreters to detect subtle differences and lateral discontinuities not obvious in standard displays and caused by the tuning effect of geological beds (thickness changes) or heterogeneity (Figure 1e). Figure 2 shows the spectral decomposition methodology in a pinch-out synthetic model. Note that each axis represents a different variable: (a) frequency content of seismic data; (b) vertical axis represents frequency domain for the surface shown and time domain for seismic data; (c) time thickness of the wedge model increasing direction. The surface represents the first spectral peak event associated with bed tuning. Deeper the surface is thinner the layer that can be measured. Note that thinner time thickness is associated with higher frequencies in the tuning cube.

Figure 1. Diagram of the spectral decomposition methodology: (a, b) following horizon interpretation, a sub-volume of the entire population of 3D seismic data is selected (stratigraphic sequence) and this sub- volume is flattened; (c, d) subsequently by the application of DFT a spectral balance is performed so that different frequencies slices can be compared; (e) the results of this processing is a tuning cube in which constant z values represent amplitude maps produced by the tuning of seismic-stratigraphic and seismic-structural features. The interpretation can be performed through animation of the constant slices in the Z direction, namely in frequency slices.

Data Segmentation Tuning Cube Parameters

The time window length and spatial position were the most sensitive parameters tested with this dataset. Since the two reference horizons did not conform well to each other, we chose the two-window option as the best. The frequency range from 0 to 80 Hz tuning cube was chosen from an earlier spectral analysis where mostly noise was found above 80 Hz. This time window position gives the sharpest overall image but combines geology from several sand layers. Calibration with well logs is needed to find out if these images reveal gross interval thickness or separate lobes.

Figure 2. Spectral decomposition methodology in a pinch-out synthetic model: the deeper the surface the thinner the layer that can be measured.

l Landmark Technical Review 76

Data Analysis and Interpretation

Interpretation can be performed through animation of the cube in Z direction, namely, frequency slices. Techniques commonly used in volume visualization can be applied, for instance, optical stacking where different frequency slices can be merged. An example is illustrated in the Figure 3 where frequencies from 5 Hz to 10 Hz were combined. In this case, for each frequency the same opacity function (transparency) was applied to each frequency map and then each one was summed or stacked optically. This makes sense only on a map view. The main idea is to extract from the seismic frequency images some geological features. Figure 4 shows the histogram of amplitude of a component frequency (5 Hz). Note the frequency slices are normalized by the median value of the distribution. Normalization is necessary in order to compare the amplitude anomalies from different discrete frequencies. We also performed a seismic-facies color code analysis based on quartiles of the distribution of tuning anomalies (Figure 5) from 5 Hz to 80 Hz in intervals of 5 Hz. Figure 5 shows a map representing the distribution of quartiles: minimum values to first quartile (black); first quartile to median (blue); median to 3rd quartile (orange) and 3rd quartile to maximum (red).
Figure 4. a) histogram curve for 5 Hz. b) each frequency slice is normalized according to its median amplitude-note that Q2 iqual to zero. Normalizing the tuning cube (Partyka 1999) Figure 3. Merging Frequency Slices Applying 3D visualization technology, this figure represents an optical stacking of difference frequency slices. ie, for each frequency the same opacity function was applied and summation of these images produced. In this case a summation of amplitudes from 5 to 80 Hz.

Application
This method is being applied in different Petrobras deep water oil fields including onshore and offshore cases, such as Campos, Espirito Santo, Reconcavo and Potiguar Basins. The figures in this article represent images from an offshore deep-water oilfield in the Campos basin.

Figure 5. Distribution analysis of 5 Hz frequency slice: Maps of quartiles: minimum values to first quartile (black); first quartile to median (blue); median to 3rd quartile (orange) and 3rd quartile to maximum (red). Red means interference produced for thicker parts and black low or no interference.

Landmark Technical Review 77

Summary and Conclusions

Spectral decomposition is an endeavor to reduce the impact of misinterpreted seismic caused by the tuning effect and/or the lack of some geological features due to seismic filtering. It provides an opportunity for a different approach to the geological interpretation of seismic information. Images obtained exhibit more geological fidelity than those produced using conventional amplitude maps. Frequently, depending on the filtering used during seismic processing, images with geological features can be interpreted up to 1/4th above the dominant frequency. With spectral decomposition, especially if the signal random noise ratio is good enough, higher frequency slices can contain useful geological information. In deep water field development, the seismic interpretation of the continuity of geological bodies is crucial for optimizing the position of producer and injector wells. If some geological features of geological discontinuities between the producer and injector wells were hidden by the standard seismic amplitude images, the risk of reduced production

and/or injectivity rate would increase. As a consequence, the economic impact of the project would be damaged. Besides the discrete frequency components available for seismic interpretation, this method also allows a set of new attributes associated with each frequency. These new attributes could be used as input for seismic pattern recognition to identify the seismic facies of the reservoirs under geological characterization.

Marple, S. L., 1982, Frequency resolution of Fourier and maximum entropy spectral estimates: Geophysics, vol. 47, 1303-1307 Partyka, G.A., Gridley, J.M., and Lopez, J. 1999, Interpretational Applications of Spectral Decomposition in Reservoir Characterization, The Leading Edge, vol. 18, No.3, pg 353-360. Partyka and Gridley, 1997, Interpretational Aspects of Spectral Decomposition, Abstract, Istanbul97. Partyka, G., Bottjer, R. and Peyton, L., 1998, Interpretation of incised valleys using new 3-D seismic techniques: A case history using spectral decomposition and coherency: The Leading Edge, 1294-1298. Widess, 1973, How Thin is a Thin Bed? GEOPHYSICS, SEG

References
Bahorich, M., Motsch, A. Laughlin, K., 2001, Amplitude responses image reservoir: Harts E&P, 59-62. Bracewell, 1965, The Fourier Transform and its Applications, McGraw-Hill. Dilay and Eastwood, 1995, Spectral analysis applied to seismic monitoring of thermal recovery, TLE, SEG. Kallweitt, R.S. and Wood, L.C, 1982, The Limits of Resolution of Zero-Phase Wavelets, Geophysics, vol.47, pg 1035-1046. Lopez et al., 1997, Identification of deltaic facies with 3-D seismic coherency and spectral decomposition cube, Abstract, Istanbul 97.

Acknowledgments
The authors would like to thank Petrobras for permission to publish this paper as well as Kenny Laughlin for their precious discussion, as well as Gene Johnson from Landmark Graphics. I

l Landmark Technical Review 78

Harts E&P October 2003

Rational Virtualization Opens E&P Eyes

services provider. Data, information and behind a companys firewall. On the other hand, project data, where the day-to-day analysis and interpretation work is performed, may or may not be hosted behind a companys firewall. As a result of It is also becoming more common that data from commercial data vendors, The exploration and production (E&P) industry has survived the Age of Irrational Exuberance with EBusiness and entered a period of Rational Virtualization. Cambridge Energy Research Associates (CERA) defines virtualization (Digital E & P Strategies Watch, Summer 2003) as using Internet-based tools and applications to perform E&P activities with a diminishing regard for geographic and organizational boundaries. At the core of successful virtualization is a data and applications hub, which includes a physical computing center, data management infrastructure, high-end visualization software and virtual team members who collaborate in real-time analyses using Web-based tools. Landmark software and services deployed in this model include: real-time drilling operations, prospect generation, exploration in mature basins and production management of these fields.
A hub system connects users with the data they need from a variety of sources.

Maggie Montaigne Landmark Graphics Corporation

knowledge are accessed through the Web, regardless of their physical location or where the data reside in the E&P lifecycle.

this holistic approach to information, users can access information of known quality in context, bringing together disparate information and systems through a single, intuitive Web-based interface. The ultimate goal of leveraging these data and applications hubs is to reduce storage and infrastructure costs and improve both asset team performance as well as financial performance. E&P companies who are actively using this rational approach to virtualization have increased personnel productivity by over 100% and reduced drilling and production costs by 10-20%.

real-time field operations and bulk data

At the core of successful virtualization is a data and applications hub, which includes a physical computing center, data management infrastructure, high-end visualization software and virtual team members who collaborate in real-time analyses using Web-based tools.
are being hosted at a hub as a service for E&P companies. Bulk data include large volumes of prestack and post-stack seismic, well log and production data. Definitive data such as well locations, well header, lease and seismic are often stored

Business practices
Just as information is only useful in context, information technology (IT) is only useful in the context of the new business practices it enables. A recent Harvard Business Review article (June 2003),

Hubs
Data and applications hubs can be deployed either internal or external to a companys firewall. The most commonly used hub today employs a hybrid method in which portions of the data and applications infrastructure are outsourced and managed by an external

Landmark Technical Review 79

Does IT Matter?, speaks directly to the types of business practice innovations occurring in the E&P industry as a result of adopting a longer-term strategic approach to IT. IT is inherently strategic because of its indirect effects it creates possibilities and options that did not exist before. In fact, many opportunities for business-practice innovations extend beyond the walls of the enterprise to include relationships with other companies. Joint venture opportunities, from a data and applications hub approach, are a perfect example of business-practice innovations that extend beyond the walls of the E&P enterprise. Instead of evaluating these opportunities in the old-fashioned way; i.e., by securing separate office space and computer equipment, they are now evaluated remotely with the participating companies sharing the same data, applications and computing infrastructure provided by an external services provider. The asset team members of each company are able to log in remotely through the Web and access all relevant information and data needed to evaluate the prospect. Cycle times are drastically reduced and collaboration amongst all parties is greatly improved. Incremental innovations Short-term initiatives (less than 12 months) are an ideal way to test specific incremental innovations in business practices, that result from implementing new IT and data management infrastructures. Distributed

Web services is an excellent example

IT is inherently strategic because of its indirect effects it creates possibilities and options that did not exist before. In fact, many opportunities for business-practice innovations extend beyond the walls of the enterprise to include relationships with other companies.

of a short-term business practice improvement in E&P because they can lead to better integration of commercial vendor data with the interpretation environment. The rapid pace of change in information accessibility and IT advancements can be somewhat daunting. The Economist (August 2-8) announced that holographic data storage would be available commercially next year in which one terabyte (1,000 gigabytes) of data can be stored on a CD. Holographic data storage will most likely be particularly useful for archival data. Innovations such as holographic data storage will someday reap huge benefits for the energy industry because it is so data and information intensive. However, in the near term, given our past experience with irrational exuberance, it is wise to focus on incremental innovations in business practices and take a more rational approach to the virtualization of the E&P industry. I

l Landmark Technical Review 80