The increased availability and accuracy of multi-sources data enhance the simulation quality of natural disasters (hazard). Moreover, it enables a better prediction of their impact on the territory (vulnerability). Numerical representation of relief (DTM) is a prime necessity in risk simulation, in particular in flood study. Integrating 2D objects into a DTM significantly improves the relief representation around each object. The aim of this paper is first to enrich the relief accuracy with the altimetric attributes of 2D vector objects, and then to assess the impact of these modifications within the context of a rise in the water level and its consequences on the road network vulnerability. The first part focuses on the integration of 2D vector data (e.g. roads) on a 2.5D digital terrain model. The integration of 2D vector data on a DTM usually raises data consistance issues. These data often stem from different acquisition modes ; moreover, their scale and their dimension (2D, 2.5D, 3D) vary according to their nature and the data capture. In order to overcome these problems, our approach consists in converting 2D vector data in 2.5D vector data by assigning them a width and computing their elevation. Then, these latter are integrated into the DTM and make it sharper, especially on the former interpolated areas. In a second part, we analyze the floods effects on the running of the road network in Orléans (France). In addition to the direct damages caused by the physical action of the water, the flood also leads to functional disruptions on the road network by forcing users to take detours. In a risk preventive viewpoint, the network administrator has first to consider a given flood scenario and consequently to identify the network components to be protected as a priority, in order to reduce to the minimum the costs induced by the detours. On account of both levels of precision considered in the original DTM and in the enriched DTM, we have to compare two damaging scenarios of the road network for a given flood hazard. To that end, we quantify the functionality of the network components with centrality measures and we compare the efficiency of the different configurations of damaged network with accessibility measures. The results of this study prove how much the flood of the road network, the induced functional disruptions and the associated preventive actions depend on the adopted precision in the representation of the terrain.