Un modèle multi-agents pour évaluer la vulnérabilité aux inondations : le cas des villages aux alentours du Fleuve Fiherenana (Madagascar)

Abstract

Natural disasters are frequent in the South West of Madagascar, particularly flooding. Assessing population vulnerability is of major importance. Vulnerability is a theoretical concept; it is not easy to assign it with a numerical value. However, there are several methods to 'measure' vulnerability to flooding. The most traditional approach is the use of an indicator that is based on the arrangement of several criteria leading on a synthetic index. In this work, we want to add a dynamical aspect to that indicator, using a modeling approach called “multi-agent system” (MAS) which is increasingly prized by geographers. The behavior of all entities brought into play (inhabitants, houses, water flows) during a catastrophic event will be simulated by a model using the agent, i.e. an autonomous individual who acts according to specific rules. Several scenarios are taken into account in order to get a dynamic indicator. This approach allows an intuitive observation of every catastrophic event and gives flexibility to add another dimension to the indicator, the forward-looking aspect of vulnerability that a static indicator cannot account for. In the long run, this type of approach can lead to a predictive model.

Keywords

• vulnerability
• multi-agent system
• flood