One of the major challenges today is to better understand and anticipate both individual and collective human behaviors face to threats and disasters, whatever their causes. This anticipation is the critical point for populations and operational services in charge of managing the crisis, since it helps ensure populations’ safety and keep the emergency situation under control as well as its spatial and temporal spread. This paper deals with the modeling of the temporal dynamics of human behaviors during a catastrophic event. Several complex models, at different scales, can be found in the literature in order to model crowd dynamics. Up to now, the models essentially take into account the panic since it is the most feared reaction. However, human sciences learn us that the population during a catastrophic event can exhibit different concurrent reactions, as behavioral sequences, and each individual does not keep the same behavior. A first work associating geographers, mathematicians, computer scientists has been proposed in this direction (Verdière et al., 2014; Provitolo et al., 2015). They have developed a first SIRbased model considering three different types of collective behaviors in catastrophic event and their different interactions. In this paper, we propose a more mature version of the mathematical model (called PCR for « Panic / Controlled and Reflex behaviors) simulating the behavioral sequences from a first phase called « reflex » to a second phase called « rational » according to the type of the threat, domino-effects and local environment.