Determining the unified representative element volume (REV) is the primary issue to be solved when studying the meso-scale pore structure of soils. We use the synchrotron radiation μ-CT

in 6.5 μm voxel resolution

to scan a Nanjing silty-fine sand specimen. Five groups of cubic pore REV with the same dimension are chosen from the three-dimensional reconstruction model

at different heights

of the specimen. The maximal ball algorithm is used to analyze each pore REV. The pore network model of each REV is developed to extract eight pore structure parameters

namely porosity

pore number per unit volume

mean volume of pore

minimum pore volume

maximal pore radius

minimal pore radius

mean radius of pore and average shape factor of pore area. Subsequently

the correlations between the eight pore structure parameters and the REV size are established respectively. The results show that all the pore structure parameters tend to converge when the REV size increases. The T test and F test are performed with different REV edge lengths for the eight pore structure parameters in sequence. The REV edge length of each pore structure parameters is determined and the maximal value of them is selected. Finally

unified REV edge length is determined at 400 voxels

namely 2.60 mm

for all the pore structure parameters. This determination method can be used for the meso-scale pore structure analysis of grain soils such as sand soil and silty soil.