Vehicular Ad­hoc NETworks (VANETs) are mainly evaluated through simulations in which the choice of a realistic wireless channel model is a central point. Deterministic channel models bring good realism but need huge simulation time, whereas with statistical models the computational effort is reduced, but sadly so is the realism of the model. In this paper, we present a semi­deterministic channel model, called UM­CRT, based on a deterministic channel simulator, CRT (Communication Ray Tracer) and a statistical channel model, SCME-UM (Spatial Channel Model Extended - Urban Microcell). To integrate it into the NS­2 network simulator, we couple it to self­developed fully compliant 802.11p and 802.11n physical layers. Simulations in urban environment show both a good realism and a reduced computation time indicating that UM­CRT is adapted for VANETs simulations.