Air pollution has become a major issue of modern megalopo-lis, where the majority of world population lives. Measuring air pollution levels is an important step in designing and assessing air quality related public policies. Unfortunately, existing solutions are inadequate to get insights on the real exposition of citizens. In particular, high quality sensors deployed today are too large and too costly to envision a three dimensional deployment at the scale of a street. In this paper, we investigate the deployment of wireless sensor networks (WSN) used for building a three-dimensional mapping of pollution concentrations. We consider in our simulations a 3D model of air pollution dispersion based on real experiments performed in wind tunnels emulating the pollution emitted by a steady state traffic flow in a typical street canyon. Our contribution is to analyze the performances of different 3D WSN topologies in terms of the trade-off between the economical cost of the infrastructure and the quality of the reconstructed air pollution mapping.