Complex Lithofacies Discrimination and Porosity Prediction of Carbonate Reservoirs Through Simultaneous Pre-Stack Seismic Inversion and Bayesian Classification: A Field Case Study of Onshore Abu Dhabi

Abstract

Abstract Much of the research has suggested that there is a clear relationship between porosity and acoustic impedance in carbonate reservoirs, being the main objective of reservoir characterization using seismic amplitudes. However, this relationship can be particularly complicated in the setting with presence of various lithofacies and different elastic responses, which cannot be completely distinguished simply by acoustic impedance. This study entailed the discrimination and occurrence probability prediction of different lithofacies through the integration of simultaneous pre-stack seismic inversion and Bayesian-based classification. Consequently, the final estimation of porosity was obtained through the integration of lithofacies occurrence probability and the lithofacies-variant linear regressions between porosity and acoustic impedance. An adequate understanding of the diverse lithofacies distribution and their different elastic properties responses to reservoir properties is essential for an appropriate reservoir characterization. Firstly, a suitable petroelastic rock physics analysis study was executed to determine the possibility of discriminating different lithofacies and predicting different reservoir properties from elastic properties, and which elastic properties act best. The four lithofacies classifications (oolitic grainstones, pack/wackestones, mudstones and dolomites) in the target reservoir were distinguished based on the cross plot between compressional wave and shear wave velocity ratio (Vp/Vs) and acoustic impedance (Zp). For this reason, a simultaneous pre-stack seismic inversion was executed to obtain the elastic properties previously identified as discriminants (Vp/Vs and Zp), which were accordingly used in a supervised classification approach through the application of Bayesian-based probability density functions computed from lithofacies information at wells. This classification scheme delivered the occurrence probability cubes related to each lithofacies and the most probable lithofacies. At the end of the process, a pseudo porosity cube was computed through the integration of the lithofacies occurrence probability (denoted as Poo, Pp/w, Pmdst and Pdol,) and the lithofacies-variant linear regressions between porosity and acoustic impedance observed at wells. The linear regressions were estimated independently for oolitic grainstones (F(ϕ,Zp)oo), pack/wackestones (F(ϕ,Zp)p/w), mudstones (F(ϕ,Zp)mdst) and dolomites (F(ϕ,Zp)dol). The resultant pseudo porosity was estimated with the following equation: Pseudo Porosity = Poo*F(ϕ,Zp)oo*Zp + Pp/w*F(ϕ,Zp)p/w*Zp + Pmdst*F(ϕ,Zp)mdst*Zp + Pdol*F(ϕ,Zp)dol*Zp. In this work, volumes of Vp/Vs and Zp consistent with variant lithofacies behaviors were obtained through pre-stack seismic inversion and served for launching a litho-seismic classification, which honors the uncertainties related to the classification approach by the probability density functions. Results have demonstrated a promising conformance with geological information and a perfect positive correlation with the well data when blind tested. Given the above, the methodology presented in this study has demonstrated the potential advantages of integrating simultaneous pre-stack seismic inversion, Bayesian-based classification and lithofacies-variant linear regressions in reservoir characterization in areas with complex lithological settings. The obtained results would be an extremely valuable and rewarding guidance and bring novel insights in further exploration and development activities in study area.