Hydraulic properties of fracture are directly related to shale-gas production, oil/gas reservoir and geothermal reservoir distribution. The permeable fracture could work as fluid pathway, but the fracture sealed by minerals (i.e. less permeable fracture) could work as seal in reservoir. Thus, hydraulic properties of the fracture are important in reservoir fluid managements. Because fracture typically has heterogeneous geometry, it is difficult to model the fracture using analytical approaches (e.g., rock physics model). Permeability of fracture is often measured by laboratory experiments, however few studies have focused on calculating permeability by using fluid flow simulation on digital fracture models. Here we used Lattice Boltzmann Method (LBM) to calculate the fluid behavior in 3D digitalized fracture models. The permeability derived from LBM simulation agrees with the laboratory-derived results. We used two natural fracture models: before and after shear deformation. We observed large-scale fluid flow network in the sheared model. We further observed permeability anisotropy for the sheared model, although the anisotropy was not clear in the non-sheared model. The permeability normal to shear direction is larger than that parallel to the shear direction. We then discussed the Representative Elementary Volume (REV) of the hydraulic properties of these fracture models. We extracted several subdomains (i.e., small fracture models) from the whole model and estimated permeability of the subdomains. When the size of subdomain is small, the values of estimated permeability widely vary by the location of subdomain. Convergence of the permeability data is found when the model size is close to the whole model. The REV of the sheared model is larger than that of the non-sheared model. Because the hydraulic properties of fracture models smaller than REV are largely influenced by local heterogeneity, the estimation of REV using the proposed method is important to calculate meaningful hydraulic properties.