Permeability Prediction Using Probabilistic Neural Network (PNN): Application to the Paleozoic Shallow Marine Sandstone of Quwarah Member, Qasim Formation, Saudi Arabia

Abstract

Abstract This outcrop analog study was conducted on surface equivalent to the Quwarah member of the middle to late Ordovician of Qasim Formation in central Saudi Arabia. The Paleozoic section contains important oil and gas reservoirs with more to explored and developed mainly related to unconventional tight gas and shale gas. The main objectives of this work is to use the probabilistic Neural network (PNN) to predict permeability of the Quwarah sandstone on the basis of systematically collected and petrographically estimated textural and compositional data from the outcrop sections of the Quwarah member. The results show that probabilistic neural network (PNN) was capable of reproducing permeability with very high accuracy, so that the calculated correlation coefficient for permeability was 0.89. This outcrop analog study, when integrated with subsurface data, might provide database, reveals heterogeneity and enhances understanding and better prediction of reservoir quality in the subsurface. 1. Introduction This paper used outcrop analog to investigate permeability variation within shallow marine sandstone of Quwarah member of the middle to late Ordovician Qasim Formation in central Saudi Arabia. The Quwarah member sandstone was deposited in shallow marine environment. The Quwarah sandstone and other stratigraphically equivalent strata are considered as targets for gas reservoirs (McGillivary and Husseini, 1992; Konert et al., 2001). However, exploration work indicated that identification of reservoir zones in the subsurface has been a challenge. This is attributed to complexity of facies, depositional environments and paleogeographic setting (Al-Mahmoud and Ibrahim 2010; Briner et al.,2010). The excellently exposed outcrop of Quwarah may provide good opportunity to examine the vertical and lateral distribution of facies and their stacking pattern at high resolution scale unavailable from subsurface data. This will help to provide better understanding of sedimentary heterogeneity and their impact on reservoir properties, quality and architecture. The main objective of this study, based on detailed outcrop analog collected high resolution data, is to predict the permeability of Quwarah sandstone using the probabilistic neural (PNN). The expected outcomes of this work is that the results might help to fill the information gap related to permeability distribution at high resolution scale, with depositional and diagenetic geological constrains taken into consideration. Outcrop studies of suitable analogue may provide more details about orientation, size, and geometry of sandstone bodies allowing more geologically realistic reservoir models to be developed (Walton et al., 1986; Hearn and Flower, 1986). This study uses sandstone reservoir outcrop analog so as predict the permeability and this will help a lot in pre-drill decisions and essential for understanding the scale and geological controls on sedimentological heterogeneity and the behavior of deeply buried sandstone reservoir (Bryant and Flint, 1993; Abdulkadir et al., 2010;Grammer et al., 2004).