“…Many efforts, including machine learning, have been proposed in the literature to improve the accuracy and applicability of cubic law. Examples include 1) modification of the definition of the aperture used in cubic law, such as arithmetic average (Brown, 1987), geometric average (Renshaw, 1995), averages proposed by others (e.g., Neuzil & Tracy, 1981;Zimmerman et al, 1991); 2) inclusion of correction factors that account for roughness influence (e.g., Louis, 1969;Patir & Cheng, 1978), tortuosity influence (e.g., Brown, 1987;Waite et al, 1999), combined roughness-tortuosity influence (e.g., He, Sinan, et al, 2021;Xiao et al, 2013); 3) implementation of machine learning techniques Sun et al, 2020). All above modifications, however, are designed for 2D rock fractures or 3D fullyopen fracture cases (e.g., no contact areas).…”