ABSTRACT: Reservoir characterization by seismic trace shape classification: Merah Besar Field, Indonesia
Bozkurt, Gokay, Keith Wrolstad, and Eko Lumadyo , Unocal Corporation, Sugar Land, TX
Variations in the physical characteristics of the subsurface are contained in the seismic signal and often revealed as changes in its shape. Therefore, understanding and recognizing the variability in the seismic character is critical for reservoir characterization. This can be accomplished by first identifying a set of traces representing the existing variability in the trace character and subsequently searching the seismic data for similarity to one of the traces in the group.
We present three approaches to constructing a group of 'seed traces' which capture the variability that exists in the data. These are (1) The Neural Network Approach, (2) Well-based Approach, and (3) The Modeling Approach. All three methods are followed by the classification step where the 'seed traces' are matched to the actual seismic data using maximum correlation and color coded with the matching seed trace color. The resulting map is a similarity map or a seismic facies map where same colors indicate higher similarities in trace shapes. These maps are qualitative for the Neural Network approach which requires minimum input for analysis (i.e., horizons). Latter two provide quantitative estimation of critical reservoir properties such as porosity, saturation and thickness by incorporating actual seismic responses at the well locations or generating synthetics by forward elastic modeling.
We have demonstrated that these methods successfully map residual gas versus commercial gas distributions, as well as pay thickness in the Merah Besar Field. These approaches have predictive capability for pre-drill assessment and provide results which can be important in risking prospects.
AAPG Search and Discovery Article #90913©2000 AAPG International Conference and Exhibition, Bali, Indonesia