Flammable and toxic gas dispersion is a key issue when designing the potential consequences of industrial accident. While resulting distances for land use planning are important, models used nowadays are quite simple. It should be reminded that several levels of approach could be used for the cloud dispersion modelling. The simplest is Gaussian models based on dispersion coefficients fitted using experimental data. Integral models enable the user to model this phenomenon quite more precisely using a simplified 1D solution of fluid flow equation in the vicinity of the release, then, for larger distances, those models used a Gaussian like approach. For those two families of model, obstacles are represented using a global ground roughness. For configurations with distributed non-regular obstacles, 3D models, as CFD or LPDM, suit better. Those models enable taking into account the different obstacles with their real characteristics and their distribution on the ground. Using those models let the user able to capture specific phenomena as for example turbulence induced by the obstacles and resulting mixture and canyon effect that can reduce dispersion. While those 3D models are more and more required regarding the importance of consequence area, two main issues remain. First, very few data are available for validating those codes in an urban like configuration. Most of large-scale experimental campaigns were achieved in free filed configuration or with regularly distributed obstacles, configuration that is relevant for Gaussian or integral validation but not for pointing out the interest and specificity of 3D models. Data available in urban like area are based on small scale wind tunnel experiments with the obvious scale effect limitation. Furthermore, since few experiments were achieved in such a configuration, some physical phenomena as the interaction between atmospheric boundary layer (ABL) and large obstacles were not studied in detail. Then, after a description of this state of the art, this paper presents an experimental campaign that aims to provide reference experimental measurement in urban like area at large scale. This campaign will also offer the opportunity in studying some specific phenomena concerning the interaction of ABL with an urban geometry.