This paper reviews and compares three recently published approaches for simplicity, validity, and parameter sensitivity. The first two approaches are based on deterministic models; the third approach uses a response equation technique in which models are defined by means of tool-response equations and interpretation-constraint equations. The first approach assumes that the weight fraction total organic carbon (TOC) is available from an external source, the rock grain density is known, and total water saturation is constant. This enables a solution for total porosity with a single equation based upon the bulk-density log, from which the gas-filled porosity can be obtained, assuming a constant water saturation. The second approach assumes that the formation consists of two constituents: porous mineral matrix and porous kerogen. It makes use of the fact that in gas shale, kerogen generally contains oil-wet porosity, and imposes gas-saturated and constant kerogen porosity. The volumes of porous mineral matrix, porous mineral kerogen, and porous mineral porosity can be obtained by using the sonic and density logs, assuming that rock grain density is known and that the porous-mineral-matrix gas saturation is a constant. The assumption of constant porous mineral gas saturation can be relaxed by iteratively using the resistivity log to update the assumed value of the hydrocarbon saturation.
This paper shows that methods 1 and 2 can be replicated by using a statistical optimization technique based on log-response equations. This technique requires some simple constraints, such as constant kerogen porosity or constant gas saturation. Moreover, the constant saturation assumptions can be easily removed, and it is possible to calibrate to core grain density and gas-filled porosity with or without wireline geochemical data, TOC, or X-ray diffraction (XRD) mineral data.
We present the log-interpretation results from using the three different approaches on a Haynesville gasshale well. The explicit models of methods 1 and 2 are tested for consistency with core data. Although method 2 suggests using TOC to establish a correlation with pyrite to improve prediction of grain density, this paper demonstrates that geochemical logs generally provide a more robust prediction of pyrite because they measure sulfur directly.
