Abstract: Large-scale water conservancy and hydropower projects are mostly built on highly steep fractured rock slope,and the seepage control structure of the slope has become a key research subject. In order to further study the permeability characteristics of fractured rock mass and the influencing factors of its anisotropy,considering the proba- bility distribution of different trace lengths and densities,the random cracks are generated by Monte Carlo method. After that the random cracks are imported into the finite element seepage calculation software RFPA2D -Flow to form a seepage fracture network model. Simulated calculation of equivalent permeability coefficient in different directions and sizes could obtain the permeability tensor in different working conditions. Combining the approximate tensor property and approximate constant property as the criterion to determine the seepage characterization unit. Also,the permeabili- ty anisotropy coefficient could be customized to study the influences of different trace lengths and densities onto the permeability tensor,the characterized element sizes and the anisotropy coefficient. The results showed that:① the den- sity and length of fractures have great influences on the principal permeability value of fractured rock mass,but have little influence on the principal direction of permeability. ② With the increase of density or trace length,the principal permeability value of fractured rock mass increases gradually. Within a certain range,with the increase of fracture den- sity or trace length,the size of characterization unit of fractured rock mass decreases gradually. ③ The anisotropic coef- ficient of permeability characteristics r in fractured rock mass decreases with the increase of density,and decreases first and then increases with the increase of fracture trace length. Based on these results,this paper took the fractured rock mass on the left bank slope of Jinping I Hydropower Station Dam Area as an example to perform the seepage numerical simulated calculation based on fracture probability distribution characteristics and related parameters obtained from ge- ological exploration. The permeability tensor and the size of characterization unit were determined from calculation,and were corrected based on on-site water pressure test results. The correction coefficient is in one order of magnitude, which shows that the simulation results in this paper are close to the measured ones.