This article deals with the social behavior modeling of a particular forest pest insect: the bark beetle. This ant-like insect has been responsible for the devastation of acres of pines trees in North America since 2005. Any tactic of forest pest management requiring prediction of pest population change over time and/or space, a realistic modeling of beetle colonies behavior would be a real benefit. The originality of this work is to propose a reactive Multi-Agent- System integrating physical diffusion phenomena. The main idea is to take into account the natural vanishing of the trail markers emitted both by decomposing trees (ethanol) and the agents that have found a source of food (pheromone). The proposed experiments show, on the one hand, that the MAS-PDE modeling leads to a realistic global behavior of the colony when considering a usual foraging scenario and, on the other hand, that, when compared with a simple reactive agent, the proposed model has a faster convergence to the asymptotic usual expected "S-shape" behavior of the agents' colony.