Lithofacies and fluid prediction of a sandstone reservoir using pre-stack inversion and non-parametric statistical classification: A case study

Abstract: This paper describes a case study that converts pre-stack seismic data into meaningful rock properties by employing non-parametric probability density functions through a probabilistic modelling approach. This study used the simultaneous pre-stack inversion method to transform pre-stack seismic data into seismic attributes like compressional impedance, shear impedance, density, and VP/VS ratio. Then cross plot analysis was conducted on selected wireline log data to identify reservoir lithofacies zones based on the ranges of properties like P-impedance and VP/VS ratio. Hydrocarbon zone was identified with the range of VP/VS ratio between 1.15 and 1.82 and ZP from 3800 to 12400 ((m/s) × (g/cc)). Water bearing sand zone was separated with VP/VS ratio with 1.85–2.12 and ZP with 3500–14900 ((m/s) × (g/cc)), and 3500–14900 ((m/s) × (g/cc)) of ZP, VP/VS ratio between 2.14 and 3.1 was used to characterize the shale zone. A non-parametric kernel density estimator is used on cross-plot data points to generate a probability density function for each lithofacies. These non-parametric PDFs were incorporated with seismic attributes using a probabilistic modelling approach based on Bayes' classification to generate a lithofacies model. The application of methodology provides a better insight into predicting and discriminating lithofacies in the study area. Highlights: Applied seismic inversion to obtain seismic elastic attributes such as compressional impedance (ZP), shear impedance (ZS), VP/VS ratio, and density.Shale, water-bearing zone, and hydrocarbon zone were identified using the cross plot analysis of well log data.Probability density functions (PDFs) for lithologies were generated on cross-plot space using the non-parameter statistical classification.Finally, hydrocarbon zones were identified using the Bayes' rule by combining the seismic data with PDFs.