We consider eight data sets obtained from synchrotron computed microtomographic experiments performed on natural and synthetic sandstones at the GSECARS 13-BM beam line at the Advanced Photon Source. Four data sets were collected on glass bead packs sintered to differing bulk densities; each of these data sets contains just over 225 million voxels (661 ¿ 661 ¿ 517), with voxel resolution of 3.34 ¿m. The other four data sets were collected from a single sample of natural sandstone. For the natural sandstone, three data sets were collected from different subvolumes within the sample, with these three data sets identical in size and resolution to those acquired for the sintered glass bead packs. A fourth data set was collected for one of the same subvolumes but at twice the resolution; this data set contains just over 1.6 billion voxels (1327 ¿ 1327 ¿ 920), with voxel resolution of 1.67 ¿m. Following processing to extract binary descriptions of the two-phase (solid and pore) structure, the image data are analyzed to calculate porosity and specific surface area. The binarized image data are applied directly in massively parallel numerical simulations of single-phase fluid flow performed using the lattice Boltzmann method; development of novel boundary conditions to facilitate efficient handling of these very large data sets is also described. The permeabilities predicted from 1 to 3 mm3 image volumes characterized at a length scale of ~1.7--3.4 ¿m are compared to experimental measurements of intrinsic permeability performed at bench conditions on standard core-sized samples (25.4 mm in diameter ¿ 50.8 mm in length, or ~26 cm3). The macroscopic permeability computed from the image data agree remarkably well with the experimentally measured bulk permeability over several orders of magnitude range in permeability and equate to over 4 orders of magnitude scale-up. The results provide insight on the representative volume and length scales necessary to characterize geometrically complex porous media and predict fluid transport properties at the macroscale.