Nearly all the members of adult population in major developed countries transport a GSM/UMTS mobile terminal which, besides its communication purpose, can be seen as a mobility sensor, i.e. an electronic individual tag. The temporal and spatial movements of these mobile tags being recorded allows their flows to be analyzed without placing costly ad hoc sensors and represents a great potential for road traffic analysis, forecasting, real time monitoring and, ultimately, for the analysis and the detection of events and processes besides the traffic domain as well. In this paper a model which integrates mobility constraints with cellular networks data flow is proposed in order to infer the flow of users in the underlying mobility infrastructure. An adaptive flow estimation technique is used to refine the flow analysis when the complexity of the mobility network increases. The inference process uses anonymized temporal series of cell handovers which meet privacy and scalability requirements. The integrated model has been successfully experimented in the domain of car accident detection.