To quantify the roughness of natural rock fracture surfaces, a two dimensional version of the modified divider method was adopted. The parameter <italic> D</italic>_r2d × <italic>C</italic>_x was found to be suitable to quantify the roughness of natural rock fractures. In addition to the mean aperture, a modified 3D box counting method was used to quantify aperture distributions of the same fractures. The modified 3D box counting method produced fractal dimensions in the range 2.3104 to 2.5661. The following new functional relations were developed for aperture parameters: (a) power-functionally decreasing mean aperture with increasing normal stress, (b) power-functionally decreasing 3D box fractal dimension with increasing normal stress, (c) linearly increasing mean aperture with increasing 3D box fractal dimension, (d) linearly decreasing mean aperture with increasing fracture closure, and (e) linearly decreasing 3D box fractal dimension with increasing fracture closure. Fluid flow through nine natural single rock fractures was measured at different normal stresses. The flow calculated for three out of the nine fractures according to sample scale cubic law using mean apertures overestimated the experimental flow by 2.2&sim;235.0 times within a normal stress range of 0&sim;8 MPa. The elementally applied cubic law (EACL) through a finite element model (FEM) also overestimated the experimental flow by 1.9&sim;111.7 times within the same normal stress range. As the normal stress applied on a natural rock fracture increases, the overestimation increases due to increasing contact areas and increasing tortuous behavior of flow. These findings clearly show the inapplicability of the cubic law to estimate flow through natural rock fractures especially under high normal stresses. New hyperbolic functions were developed to relate mean aperture to the power <italic>n</italic> to applied normal stress at both the sample and finite element scales. The following new functional relations were developed between fluid flow rate and the aperture parameters: (a) power-functionally increasing flow rate per unit head with increasing mean aperture, (b) exponentially decreasing flow rate per unit head with increasing fracture closure, and (c) power-functionally increasing flow rate per unit head with increasing 3D box fractal dimension.