Abstract: In order to characterize the mesoscopic seepage of pore spaces within undisturbed granite residual soil (UGRS). The 2D/3D pore models of UGRS were reconstructed using computed tomography (CT) scanning and image processing technology. The lattice Boltzmann method (LBM) was employed to simulate the seepage displacement process and water distribution in UGRS pores through programming. Results indicate that isolated pores have minimal impact on seepage outcomes. Only a few large, straight pores with good connectivity become dominant channels, generating preferential flow. Throughout the seepage process, the maximum velocity in the dominant passage shifts from the inlet to the middle and then to the outlet region. The influence of pore structure characteristics on seepage is more significant than pore proportion. Using industrial CT scanning technology and LBM provides a better description of real soil pore flow. The 2D model intuitively illustrates the preferential flow phenomenon of macropores, suitable for qualitative research. The 3D model, closer to reality, accurately calculates soil permeability, making it suitable for quantitative research.