Borehole radar is known as a powerful technique for monitoring of subsurface structures such as water flow. However, conventional borehole radar systems are operated in the frequency range lower than 100 MHz and the resolution is poor to measure a surface roughness and an inner structure of subsurface fractures directly. In order to monitor the water flow, these characteristics of subsurface fractures are important. We developed a polarimetric borehole radar system using dipole antennas and axial slot antennas and have found that this system can provide more information than conventional borehole radar. However, the relationship between the characteristic of subsurface fracture and the measured polarimetric radar information has not been clear. In this paper, we simulate electromagnetic wave scattering from subsurface fractures having a rough surface by Finite-Difference Time-Domain (FDTD) technique and discuss the relationship between a surface roughness of subsurface fracture and the polarimetric information. It is found that the subsurface fracture having strong cross-polarized components can be estimated to be rough surface fracture. The full polarimetric single-hole radar measurement was carried out at the Mirror Lake site, NH, USA. In this experiment, we found that subsurface fractures can be classified into some groups by an energy scattering matrix, and found that the subsurface fracture estimated to have a rough surface corresponds to that has higher water permeability.