Abstract:To eliminate the non-physical reflection at the truncated boundary of GPR forward modeling, it is necessary to deal with the simulated region boundary. At present, the well applied method is to load the perfectly matched layer, and the complex frequency shift perfectly matched layer (CFS-PML), which can enhance the absorption effect of litter wave and prosperous lost wave in low frequency region and suppress the false reflection of signals, is widely used in GPR forward modeling. In this paper, CFS-PML is applied to the numerical simulation of solving the two-dimensional GPR wave equation with the symplectic partitioned Runge-Kutta (SPRK) algorithm, of which the symplectic algorithm can be utilized to realize the symplectic calculation of the time and space coordinates, so as to accomplish the high-precision forward modeling of GPR. Firstly, taking the TM wave as an example, we deduce the SPRK of GPR wave equation solution formula based on CFS-PML. Then, a narrow and long model is utilized to conduct the experiment of the key parameters’ selections of CFS-PML absorbing boundary condition, and the optimal parameters were determined by comparing the reflection error. Moreover, the comparison between CFS-PML and transmission boundary shows that the absorption effect of CFS-PML boundary condition at large Angle grazing is better than that of transmission boundary. Finally, to further verify the effectiveness of our proposed algorithm, the SPRK program loaded with the CFS-PML boundary condition is conduct on a geoelectric model with undulating topography. Then the forward GPR cross-sectional view, which is attained by using both cross-sectional method and wide-angle method, proved that the algorithm can fit the uneven underground interface well.