Abstract: Abstract: Surface crusts are often hardened with the platy surface of soil layers, particularly widely distributed in arid and semi-arid areas. There are also some important influences on the soil and water conservation, as well as the ecological environment, due mainly to the special pore structure and hydraulic characteristics. In this study, three-dimensional images of real soil crust samples were acquired to characterize pore morphology and structure distribution, combining X-ray tomography with pore-scale simulation. The water infiltration process was then simulated in soil samples, thereby to quantitatively analyze the relationship between pore structure and hydraulic characteristics. Three soil samples were collected with the surface crust (the porosity of 0.196, 0.144, and 0.093) using X-ray tomography. Furthermore, a systematic investigation was made on the variation of pore structure along the depth direction of crust soil samples. Specifically, each soil sample was then sliced into 9 layers from the upper surface with each layer of 100 voxels depth. The permeability of each soil layer was then calculated using the lattice Boltzmann method. Subsequently, the correlation analysis was conducted between structural characteristic parameters and permeability. A fractal dimension model was also proposed to calculate the permeability using soil porosity and complexity. The pore structure analysis showed that the porosity and pore connectivity of the surface and bottom layer were smaller than those of the intermediate soil layer in the three soil samples. The maximum pore connectivities of soil layers were 99.21%, 94.64%, and 57.35% in the three soil samples located in the middle of the soil, respectively. The minimum were 45.99%, 27.30%, and 11.74%, respectively, located at both ends of the soil sample. The pore structure of topsoil was different from that of the adjacent soil layer, where the variation coefficient was E>5%. The average pore diameter of each soil layer decreased, with the increase of depth, from 0.43, 0.37, and 0.50 mm to 0.15, 0.14, and 0.14 mm, respectively. More importantly, the soil layer with higher porosity presented better pore connectivity and lower pore complexity, where the correlation coefficient was 0.972. The permeability calculation showed that the surface layer of soil samples behaved low local permeability of 7.472, 0.006, and 6.960 μm2 in the three samples, respectively, indicating there was a blocking effect of crust on soil seepage. Better permeability was achieved in the soil layer with higher porosity, better pore connectivity, and lower pore complexity. As such, a new permeability prediction model was finally proposed, where the specific surface area in the Kozeny-Carman equation was replaced by pore complexity. A higher prediction accuracy was obtained in the improved model, where the determination coefficient R2 increased from 0.96 to 0.98, while the root mean square error (RMSE) decreased from 25.06 to 18.31 μm2, compared with the simplified one. This finding can offer strong support to water resources management and ecological environment protection in arid and semi-arid areas.