We develop a simulation tool which reproduces lane departure crashes for the purpose of estimating potential benefits of Lane Departure Warning (LDW) systems in the American traffic environment.Tools that allow a fast evaluation of active safety systems are useful to make better systems, more effective in the real world; however accuracy of the results is always an issue. Our proposed approach is based on developing a simulation tool that reproduces lane departure crashes, then adding the effect of the LDW to compare the cases with and without the safety system, and finally comparing the results of different settings of the safety system. Here, the accurate reproduction of the relevant crashes determines the reliability of the results.In this paper, we present the reproduction of the lane departure crashes occurred in American roads in one year, by using data distributions obtained from retrospective crash databases. We analyze data from NASS/GES1 and NASS/CDS2 to obtain the characteristics of lane departure accidents in the USA. Then, we model lane departure crash scenarios by dividing them in 5 phases: (A) before departure, (B) starting the departure, (C) running off-lane, (D) impact and (E) after the impact. This division of the departure sequence allowed an easy recognition of the required parameters to reproduce each phase. The reproduced crashes were validated by comparing them to the real crashes.The reproduction of lane departure crashes will allow the estimation of potential benefits of LDW systems in the USA, which is an important step towards the development of more effective countermeasures.