While empirical evidence shows the main effects of the driving style (principally speed choices) on safety, this study aims to quantify the influent parameters and their interactions upon the roadway departure risk. A previous work [1] using a sensitivity analysis concludes that speed, lateral position on the pavement, total embedded mass and mass center position are the relevant parameters. They constitute what is called initial conditions in the following. Probabilistic methods for the risk assessment are proposed to deal with uncertainties arising from the road infrastructure, the vehicle characteristics and the driver behaviour. Those methods originally developed in the field of structural reliability reveal promising interest in the road safety question as they allow the direct computation of a risk index, not provided by a deterministic modeling. This approach involves both measurements on real traffic by static video sensors and simulations from a specifically developed model. A set of 400 experimental trajectories is used to define mean trajectories among different classes of drivers for one turn, and simulated ones are necessary for the reliability index calculation. The result of this study is a surface response of the roadway departure probability against initial speed and lateral position on the pavement. This work will lead to the development of a warning system within the French national project SARI. Envisaged alert procedures are warning light flashes inviting the driver to modify his trajectory if the risk index calculated at the entrance of the curve indicates a dangerous situation. Difficulty for short-term implementation comes from the statistical characterization of the initial conditions (i.e. mean value and standard deviation), especially for mass and mass centre position measurement device on standard vehicles.