Within the framework of the Environmental Noise Directive (END) 2002/49/CE concerning the assessment and management of environmental noise, cities of more than 100,000 inhabitants are required to construct noise maps to be used in defining noise abatement action plans. For consistency, a preliminary evaluation (mainly by acoustic simulations) of the effectiveness and impacts of these action plans is mandatory. The use of Geographic Information System (GIS) software is a tool that has become essential for such studies. GIS offers many spatial analysis functions and cartographic capabilities that are useful for understanding the impact of action plans on noise maps at different scales (i.e., buildings, administrative units, etc.). The connection between GIS and noise prediction was established more than two decades ago in a Dutch study on the impact of road traffic noise on bird reproduction. As early as 1986, simulations were carried out using Silence software on a large area in the Netherlands. Unfortunately, little information is available on the acoustical aspect of the software. The application at hand was clearly focused on the countryside and thus has few implications for urban areas. Today, Silence 31 combines GIS with the commercial Predictor software with a focus on highways rather than urban areas. In this paper, we propose an alternative approach, implementing a noise prediction method within the OrbisGIS2 software. The noise emission and propagation models are inspired from the French national method “NMPB 2008”, but are simplified for two dimensional calculations in urban areas. The method, as well its implementation in the GIS software, are detailed in sections 1 and 2. We then present a case study in section 3 to evaluate several mobility plans in terms of noise impact. The particular interest of this approach is that we implement the model in the two-dimensional GIS software OrbisGIS and pay special attention to algorithm optimization in order to reduce computational times and resource consumption. Our method is able to produce noise maps for very large domains (around several millions of square meters) on a personal computer in just a few hours.