In sparsely fractured crystalline rock, aperture variability exhibits significant control of the flow field through the fracture network. However, its inclusion in models is hampered due to a lack of field measurements and adequate numerical representation. A model for aperture generation is developed based on self-affine methods which includes two key parameters, the Hurst exponent and a scaling parameter, and which accounts for relative anisotropy and correlation between the adjacent surfaces forming the fracture. A methodology for analysing and extracting the necessary parameters from 3D surface scans of natural rock fractures is also developed. Analysis of the Hurst exponent and scaling parameter space shows that input combinations following a linear upper bound can be used to generate aperture fields which accurately reproduce measurements. It is also shown that the Hurst and scaling parameters are more sensitive than the correlation between the upper and lower fracture surfaces. The new model can produce an aperture ensemble that closely corresponds with the aperture obtained from the surface scans, and is an improvement on previous methods. The model is also successfully used to up-scale fracture apertures based on measurements restricted to a small sub-section of the sample. Thereby, the aperture fields generated using the model are representative of natural fracture apertures and can be implemented in larger scale fracture network models, allowing for numerical simulations to included representation of aperture internal heterogeneity.