In order to clarify the relationship between the permeability of cement-based material and its pore structure, several mortars at different water to cement ratios and curing temperatures were investigated through the measurements of their pore structures by the low-field nuclear magnetic resonance (LF-NMR) and mercury intrusion porosimetry (MIP). Their permeability coefficients to water were also measured based on the Darcy law. The results show that the pore size distribution of water-saturated mortar is obviously different from that at totally dry state, whose critical diameter is larger by roughly 1 order of magnitude. This great gap is rooted to the essential water permeability for hydrated calcium silicate (C-S-H) and thus cement-based material, making their nanoscale pore structure highly depend on water contents. Furthermore, the total porosity and critical pore size both remarkably increase after curing in hot water of 80 ℃ for 56 d, which therefore increases the water permeability. These effects are brought by the chemical aging of C-S-H gels, which harms the durability performance of cement-based material. When investigating the evolution of pore structure through common MIP, the effect of drying preconditioning on pore structure should be taken into account because it is much greater than the effect of chemical aging at elevated temperature. If treating all the pores as bundles of capillary tubes, the classical Kozeny-Carman theory is able to predict the water permeability of mortars according to the pore structure achieved through LF-NMR at water-saturated state. The relative prediction error of the Kozeny-Carman theory with a linear coefficient of 1.09 is as low as within [-42.7%, 71.1%], which is similar to the experimental error of extremely low water permeability.