Assessing the prospectivity of unconventional reservoirs presents several challenges (Wang et al., 2016). These include: (1) complex geological characteristics of the target formations, (2) selection of appropriate and relevant evaluation parameters and methods, and (3) identification of intervals with high potential for resource extraction. A key step in this process is determining the most prospective zones—commonly called sweet spots—within an unconventional play. The assessment substantially reflects two parameters: total organic carbon (TOC) and brittleness. TOC represents the abundance of organic matter within sedimentary source rocks that has the potential to generate hydrocarbons under elevated temperatures, while brittleness reflects the formation’s capacity to undergo effective natural or hydraulic fracturing—an essential property for enhancing permeability in unconventional reservoirs, which are typically characterized by ultralow permeability (Iyare et al., 2021; Ore Schmoker, 1979) have been developed to estimate TOC continuously along the wellbore. More recently, artificial intelligence (AI) and machine learning (ML) techniques (e.g., Davy et al., 2024a,b) have also been applied to generate continuous TOC profiles with improved accuracy. The aforementioned study (Davy et al., 2024a) incorporated a petrophysical constraint called δ, derived from the difference between neutron porosity and bulk density, which effectively distinguished clay-rich (non-prospective) from clay-lean (prospective) intervals. A similar parameter, Δ, inspired by Hall et al. (2016) and based on the difference between neutron porosity and density porosity, exhibited comparable behavior Brittleness estimation encounters difficulties similar to TOC, particularly due to limited core data and the subjectivity in establishing reference trends that interpolate or extrapolate between sparse measurement points. The Brittleness Index (BI), often used for this purpose, is difficult to generalize from sparse core samples alone, prompting empirical methods. Mews et al. (2019) categorize brittleness estimation into three main approaches: mineralogical, log-based, and elastic-based. The mineralogical-based brittleness index (MBI) links higher quartz content to greater brittleness and higher clay content to reduced brittleness, while the effects of calcite and dolomite vary across models. The log-based brittleness index (LBI) typically uses conventional well logs (e.g., neutron porosity, gamma ray), while the elastic-based brittleness index (EBI) relies on elastic parameters such as Poisson's ratio and Young's modulus, which may be derived from either well logs (e.g., dipole sonic and density) or laboratory measurements. While this difference may overlap in practice, particularly since elastic properties may often be derived from logs, this classification calls attention to different data types and modelling assumptions. Mews et al. (2019) state that a universal brittleness model is unrealistic; therefore, they favor the MBI as the one that represents the intrinsic property of the rock’s material.
Davy et al. (Tue,) studied this question.