A new consortium hopes to combine seismic, logging and rock physics to study inelastic rock properties.

Rock Solid Images, in cooperation with Ødegaard AS and Petrophysical Consulting Inc. (PCI), has begun a project to develop, test and commercialize technology for a new type of direct hydrocarbon detection. The method will focus on the use of inelastic rock properties to greatly enhance the sensitivity of surface seismic methods to the presence of hydrocarbon saturation. This new technology is expected to be particularly valuable as a tool for natural gas exploration and development. The end result will include use of energy absorption, dispersion and attenuation (Q) along with traditional seismic attributes like velocity, impedance and amplitude variations with offset (AVO).
The technology is being disseminated in real time via a project Web site and periodic face-to-face meetings between the participants. All companies involved will be contributors to the resulting technology. Each oil company sponsor will be allowed to submit a data set for analysis using the new methodology. Twelve sponsors represent several exploration and production business sectors, including large independents, major integrated oil companies and national oil companies.
The approach is to combine three elements:
• rock physics understanding of how inelasticity is related to rock type, pore microstructure, pore fluid types and stress;
• synthetic seismic modeling that will help identify the relative contributions of scattering and intrinsic inelasticity to apparent Q attributes; and
• robust algorithms that extract relative wave attenuation attributes from seismic data.
These elements, combined with forward modeling from logs and advanced neural network classification of seismic data, will form the essential elements of a commercial technology.
Direct hydrocarbon detection
During the past two decades, several strategies have emerged for detecting hydrocarbons from seismic data. Most of these are based on rock elastic properties - travel time (or velocity), impedance, bright spots - and can be understood deterministically in terms of the compressibility and density of the pore fluids coupled with the stiffness of the rock matrix. The mechanics of these elastic fluid signatures at low seismic frequencies are described reasonably well and quite simply by the Gassmann (1951) relations. AVO analysis, which uses inference of compressional (P) and shear (S) wave impedances, helps to further separate pore fluid from lithologic effects. Newly emerged methods extract AVO-like amplitude information in the form of impedance inversions of angle stacks. Many of these hydrocarbon indicators, though initially discovered empirically, can be understood deterministically and theoretically through the science of rock physics.
In spite of these successes, practical problems of quantifying hydrocarbon indicators remain and tend to stem from two sources: the difficulties of accurately extracting attributes such as velocities and impedances from seismic data, and interpretation nonuniqueness. Velocity and impedance are certainly affected by the pore fluids we wish to detect as well as by variations in porosity, clay content, pore pressure, temperature and stress. In other words, most methods depend on estimates of no more than two or three fundamental seismic properties (P-velocity, S-velocity and density, or combinations of them), while there can be many more lithologic and fluid unknowns.
Project leaders will demonstrate novel and robust techniques for reducing hydrocarbon indicator risk by exploiting an additional set of completely independent indicators - the rock inelastic properties. All real rocks deviate from simple elasticity in the form of frequency-dependent wave velocities (velocity dispersion) and intrinsic wave attenuation. Laboratory and theoretical work dating back to the 1940s and 1950s illustrates that attenuation, in particular, can be extremely sensitive to fluid properties. Yet attenuation has seldom been applied in commercial interpretation of seismic images, owing in part to the difficulty of estimating attenuation from seismic and the lack of an efficient integrated log, rock physics and seismic analysis strategy.
Seismic wave attenuation
Seismic dispersion and attenuation as a tool for hydrocarbon exploration and exploitation has been understood in principle for many years. However, it is only becoming technically viable owing to the recent tremendous advances in seismic data quality and computing. Historically, most attempts to use inelastic rock properties in real-world geophysical applications were doomed by the inadequate state of amplitude preservation in commercial seismic surveys. However, this has been changing rapidly and will continue to advance in the coming years.
Calibration of seismic-derived attenuation from borehole-derived attenuation and dispersion measurements will be an important part of this study. Work in this area is fairly recent. This technology also has improved significantly thanks to new-generation, high-energy downhole acoustic tools.
Another key element in the application of attenuation to oil and gas exploration is the extraction of attenuation attributes from field seismic data. This was discussed as early as 1969. Many publications since have expanded and enhanced the understanding of seismic attenuation measurements in the field.
Initial results
This 30-month project began in January 2001, and the consortium has been working primarily on the rock physics fundamentals of attenuation as a hydrocarbon indicator. In this first phase of the project, the focus is to assemble a knowledge base of existing rock physics data and relations on seismic attenuation and how they relate to lithology and pore fluids.
The goal is to develop the methods to predict Q pseudo-logs. Existing well log quantities such as compressional wave velocity (Vp), shear wave velocity (Vs), porosity, shale volume (as a volume fraction or percent of solid minerals, Vshale) and water saturation (as volume fraction or percent of pore space, Sw) will be used to predict a corresponding estimate of Q vs. depth as well as perturbations to these Q estimates corresponding to gas and other hydrocarbons.
These pseudo-logs will be the key inputs to the systematic forward-modeling of synthetic seismograms from which Q attributes will be extracted and calibrated to the actual interval values. The purpose of this modeling is to establish the sensitivity of the attributes to the properties of interest and help separate the contributions of intrinsic Q, which relates to the rock and fluid properties, and scattering Q, which relates to the geometric arrangement of layers. This is viewed as a natural extension to the Q domain of the usual forward modeling studies used in any careful rock physics study of the signatures of rock and fluid properties for exploration and reservoir characterization and monitoring. It is analogous to the "what if" exercises routinely used, for example, in AVO modeling. Figure 1 shows how saturation effects can be separated on a loss diagram. Water-saturated sands (blue symbols) are characterized by high Vp/Vs and Qp/Qs values, whereas the gas sands (red symbols) have low Vp/Vs values but high Qp/Qs. This implies that the shear losses are as high as bulk losses in gas sands, whereas in saturated sands shear losses are higher.
Attenuation from seismic data
A crucial part of this project will be the extraction of quantitative attenuation, dispersion and Q information from field seismic data. Several methods are being pursued to accomplish this goal. This project focuses on use of pre- and post-stack data and will investigate methods based on vertical seismic profiling data. One key method being developed is continuous computation with Gabor-Morlet decomposition, which divides the frequency domain into a smaller number of subwindows related to the time resolution by the Uncertainty Principle. This method allows project leaders to independently estimate absorption and dispersion from the amplitude and phase spectra variation. This may become a major breakthrough in Q determination from surface seismic data. Figure 2 shows a preliminary example of an attenuation cross-section from the Gulf of Mexico using a Garbor-Morlet method.
Neural networks
Several approaches combine well log-derived information and seismic attributes for the purpose of predicting rock properties. The technique used in this project is based on a proprietary artificial neural network (ANN) and is an adaptation of the Rummelhart method that employs the delta rule with back-propagation of errors. The key to training the ANN is to provide it with a sufficient number of examples to allow the network to recognize the reservoir characteristic of interest, say hydrocarbon saturation. The multiple training cases are produced from multiple well logs and the pseudo-wells. In most instances, some data is kept out of the training data set and used for verification, a process known as "blind well" testing. A suite of seismic attributes including Q, fluid factor, AVO and acoustic impedance will be used to train the ANN to predict the hydrocarbon saturation, lithology or porosity classes at each well.
For further information about the project, including membership details, contact Dr. Joel D. Walls, project coordinator, at j.walls@rocksolidimages. com.

Comments