The utilization of the road network by vehicles with different
		behaviors can generate a danger under normal and especially under
		evacuation situations. In Le Havre agglomeration (CODAH), there are 33
		establishments classified SEVESO with high threshold. The modeling and
		assessment of the danger is useful when it intersects with the exposed
		stakes. The most important factor is people. In the literature,
		vulnerability maps are constructed to help decision makers assess the
		risk. These maps are based on several types of vulnerability:
		socio-demographic, biophysical and other different types of hazards.
		Nevertheless, such approaches remain static and do not take into
		account the population displacement in the estimation of the
		vulnerability. We propose a decision support system which consists in a
		dynamic vulnerability map based on the difficulty to evacuate different
		sectors in Le Havre agglomeration. This map is visualized using the
		Geographic Information System (GIS) of the CODAH and evolves according
		to the dynamic state of the road traffic through a detection of
		communities in a large graph. This detection is realized by an ant
		algorithm.