A prediction model of GPS multipath requires a constant trade-off between computation time and realistic modeling. In previous papers we have presented a GPS multipath simulator based on physical optics. Regarding the design of this deterministic model, we have carefully justified the choices we have made to obtain reasonable computation times. They concern both the electromagnetic method and the precision of the data used to describe the 3D scene. To account for the limits of these choices, we define in this paper a new statistical model and we propose an innovative hybrid deterministic-statistical GPS multipath simulator adapted to airport navigation. This hybrid simulator is based on the deterministic simulator via Monte-Carlo simulations. Its statistical origin is the variability of the environment. For a particular scene its outputs are the statistical moments of the GPS range error. We show the necessity of this statistical component for realistic prediction purposes in comparison to pure deterministic predictions by means of simulations on a test-case.