Optimizing warning siren placement for audibility coverage using acoustic modelling and genetic algorithms

In contexts ranging from natural disasters to technological accidents and security threats, sirens play a crucial role in alerting the population by providing a rapid and widespread warning capability. Optimizing the spatial deployment of sirens to maximize audibility for the target population remains a critical and underexplored issue. In this study, we employ open-source tools for environmental noise modelling and multiobjective optimization: NoiseModelling, based on the CNOSSOS-EU propagation framework, and OpenMole, implementing the NSGA-II evolutionary algorithm. These tools are coupled to explore the solutions space and identify Pareto-optimal configurations according to two objectives: (1) the number of buildings exposed to sound levels above 80 dB, and (2) the total area exposed above this threshold. A case study on Saint Barthelemy Island suggests that, under the modelled conditions, the optimized Pareto front ranging from 7836 to 7858 dwellings and territorial coverage ranging from 15.29 to 15.31 km 2 yields higher predicted coverage than the configuration proposed by domain experts (6658 dwellings, 12.30 km 2 ). The comparison between expert-based and model-based solutions reveals methodological limitations, such as the integration of non-acoustic contextual factors, and the strong potential of this approach as a decision-support framework for the design and evaluation of siren alert networks.