The manuscript discusses the uncertainty and complexity intrinsic to the modeling of the accident occurred in Viareggio on June 29th, 2009 where a freight train loaded with LPG derailed and one of the tank cars released 45 t of LPG that finally exploded inside some houses and burnt. Thirty-one people died, thirty citizens were seriously injured and the damages were valued at 32. M€. The paper highlights the influence of the input variables on the accident consequences to determine the most probable sequence of events. Given the set of input data for the simulation of the accident, the most important of them were varied, within a suitable interval, to assess their influence on the accident consequences in terms of time the dense-gas cloud took to reach the farthest exploded house of Terminetto neighborhood (i.e. about 100 m from the accident epicenter), and the pool-fire dimensions. The analysis reports the difficulties in applying quantitative models to realaccidents, when the events do not comply with the model hypotheses. Nonetheless, the manuscript shows that it is possible to reconstruct the most probable sequence of events by investigating the more likely scenarios. The analysis shows that the assumption of a circular pool leads to the overestimation of the pool-fire height. In addition, the time of arrival of the dense-gas cloud does not present large variations among the different accident scenarios and it is always too short to allow implementing any mitigation actions or effective emergency plans.