Classical geostatistical inversion methods perform a sequential approach where acoustic impedance (AI) values are generated in a seismic trace by trace basis (Bortolli et al, 1993, Haas A. et al, 1994, Grijalba-Cuenca et al, 2000). These methods need a priori knowledge of areas of poor seismic quality, otherwise false "good" correlations between synthetic and real seismic can be artificially imposed.
A new version of the stochastic seismic inversion algorithm (Soares et al, 2006) was recently applied to a complex Middle East carbonate reservoir to evaluate the robustness and efficiency of the new methodology. This reservoir, located onshore Abu-Dhabi, comprises a strong progradational internal geometry with several clinoforms which amplify the internal complexity of the reservoir and have a strong impact on well production.
This iterative methodology is based on global simulations and co-simulations of AI. The process generates several intermediate models. Those that contribute more to achieve an objective function are used to condition the simulations of the next iterations. There is no local imposing of artificial good fit, as it happens in most trace-by-trace approaches. Spatial dispersion and main patterns of AI are reproduced at the final AI cube. Another important advantage of this technique is that the final images reflect the quality of the seismic data through the uncertainty associated to the final seismic impedance cube.
Despite the scarcity of the log data and the small area of the pilot zone compared with the total volume of the entire seismic cube, the proposed method achieved extremely satisfactory results. Both synthetic seismic data and the seismic acoustic impedance cube captured the main local geologic features of this complex reservoir. Final local correlation coefficients, measuring the match between synthetic and real seismic, can be viewed as a measure of the local uncertainty associated to the areas of good and poor seismic.