GEOPHYSICS, VOL. 65, NO. 1 (JANUARY-FEBRUARY 2000); P. 46–53, 7 FIGS., 2 TABLES. Case History Permeability evaluation in a glauconite-rich formation in the Carnarvon Basin, Western Australia Yujin Zhang ∼ , Henry A. Salisch ∼ , and Christoph Arns ∼ ABSTRACT It is usually difficult for petroleum engineers and geo- scientists to obtain reliable estimates of permeability from geophysical logs, especially in lithologically com- plex formations such as the Mardie Greensand Forma- tion in the Carnarvon Basin, Australia, which consists of lower Cretaceous glauconite-rich sandstones. This paper presents an alternative petrophysical evaluation of per- meability in this formation through the integration of the geological and petrophysical analyses. Neural network techniques were used to establish permeability predic- tion models in cored wells or sections and to predict per- meability from well logs in uncored wells or sections. The permeabilities obtained from minipermeameter mea- surements were taken as the basis and reference for the petrophysical evaluation. Four log-derived parameters, which best reflect the permeability in the Mardie For- mation, were defined and extracted from the available conventional logs. These parameters (not original log re- sponses) were taken as the log inputs to evaluate perme- ability. Through the training, testing, and validation of the networks using the log and core data in the cored in- tervals, a permeability prediction model/network was es- tablished. Further, the permeabilities in 15 wildcat wells were determined from conventional well logs. The re- sults indicate that the petrophysical evaluation of perme- ability is valid and applicable in the Mardie Formation. INTRODUCTION Permeability is a key parameter in formation evaluation and reservoir characterisation that must be known or estimated to locate potential reservoirs. Estimation of permeability, in most Presented at 67th Annual International Meeting, Society of Exploration Geophysicists. Manuscript received by the Editor October 20, 1997; revised manuscript received May 25, 1999. ∼ School of Petroleum Engineering, University of New South Wales, Sydney, NSW 2052, Australia. E-mail: z.yujin@unsw.edu.au; c.arns@unsw.edu.au; h.salisch@unsw.edu.au. c ∼ 2000 Society of Exploration Geophysicists. All rights reserved. cases, is obtained log analysis, core analysis, and/or well testing. Frequently, permeability data from core analysis and well test- ing are not available for most wells in a given field or at every one of the required depth levels because of borehole condi- tions and/or the high cost of obtaining cores. Log analysis of- fers a unique opportunity to determine a continuous profile of permeability within a borehole. Log response data are usually available for each well (cored and uncored wells) in a given field or basin as logging measurements are economical and quick. Therefore, it is essential to accurately determine permeability from well logs. For many years attempts have been made to use wireline logs for this purpose. Much has been published on the subject, and a number of methods and models are being used to achieve this goal. Based on conventional well logs, two categories of methods to determine permeability are generally available. One category contains the so-called “standard meth- ods” based on empirical and theoretical relationships among porosity, irreducible water saturation, shale content, etc., such as the methods described by Wyllie and Rose (1950), Timur (1968), and Coates and Dumanoir (1974), as well as some modified forms of these methods. Another category embodies the so-called “statistical methods” based on statistical relation- ships described by Nicolaysen and Svendsen (1991), Sinha et al. (1994), Johnson (1994), and others. In addition, permeability can also be estimated by more modern logging techniques such as waveform sonic logging (Buffin, 1996), geochemical logging (Ahmed et al., 1989), nuclear magnetic resonance (NMR) log- ging (Howard et al., 1997), and imaging logging (Coskun and Wardlaw, 1993). The Mardie Formation is one of several productive reser- voirs in the Carnarvon Basin, Western Australia. In order to identify potential hydrocarbon reservoirs in this type of for- mation it is necessary to describe and evaluate the reservoir quality. Permeability is a key parameter to reflect reservoir quality in the Mardie Formation. The purpose of this work is 46