A Novel Approach to Reservoir Characterization Using Seismic Inversion, Rock Physics and Bayesian Classification Scheme
Nader C. Dutta
Rock-physics analysis can provide the relationship between the parameters (or seismic attributes) that govern seismic-wave propagation (e.g. Vp, Vs and density in isotropic media) and the reservoir property of interest, such as rock or fluid type, porosity, pressure and saturation. In this process, we need to account for the quality of the seismic data and derive the appropriate uncertainties associated with the seismic data, such as noise, resolution, and inversion artifacts into the reservoir property estimation. In this presentation, we show how to quantitatively propagate seismic data quality issues such as resolution, noise, and inversion accuracy into the lithology estimation in a clastic basin. The method consists of several steps: seismic inversion to obtain elastic parameters, petrophysical well-log analysis to define a classification scheme based on Bayes’ Theory and probability density functions (PDF); upscaling the PDF’s to seismic scale using Backus’ Theory and finally, applying the final scheme on seismic attributes (Vp, Vs and density) derived from the first step. The use of full-waveform inversion and Bayesian classification techniques provides a mathematical framework that enables us to model and directly relate data quality input into the uncertainty associated with reservoir properties prediction. The final output of this process is a map in 2-D and a cube in 3-D, of rock and fluid types with confidence levels associated with each property at each common mid-point (CMP) and time sample. We illustrate the procedure with examples from several clastics basins: Gulf of Mexico and India. This methodology can be easily applied to data from carbonates areas as well where inversion techniques are known to yield porosity, pay and fracture properties.
AAPG Search and Discovery Article #90077©2008 GEO 2008 Middle East Conference and Exhibition, Manama, Bahrain