This study presents the performances of an inverse modelling approach aiming in identifying position and emission rate of a localised pollutant source placed within a city district. To that purpose we combine wind tunnel experiments and an urban dispersion model. Experiments are performed in an idealised urban canopy, made up of regularly spaced blocks, and provide the pollutant concentration field downwind the source within the canopy. The urban dispersion model, named SIRANE, is an operational model that simulates the main mechanisms governing the pollutant transfer within a network of streets. INTRODUCTION The risk management of accidental atmospheric pollutant releases in a built environment is a major concern within both industrial sites and urban areas. To this purpose we need to identify rapidly the position and the strength of the pollutant source. This requires the application of inverse modelling techniques together with properly designed monitoring network. Previous authors (Lushi and Stockie, 2010; Chow et al., 2008) have combined inverse modelling techniques with different direct models, namely CFD codes and Gaussian dispersion models (Rudd et al., 2012). Both approaches show major limits for operational risk management in a built environment. CFD codes require long computation times, which are not consistent with crisis management. On the other side, the results of Gaussian model will be affected by significant errors due to an oversimplification of the velocity field, since these models are not able to simulate local effect due to obstacle wakes and streets channelling. In this study we take advantage of recent advances in urban dispersion modelling, adopting a new direct operational model, named SIRANE (Soulhac et al., 2011) in order to simulate pollutant dispersion in idealised urban geometries, such as those investigated experimentally by Garbero et al. (2010). The aim is to identify a single stationary pollution source placed in a city district and whose position and flow rate are unknown, from a varying number of direct observations of pollutant concentration. WIND TUNNEL EXPERIMENTS The experimental measurements used in this study are those presented by Garbero et al. (2010) and carried within the wind tunnel of the Laboratoire de Mécanique des Fluides et d'Acoustique of the Ecole Centrale de Lyon. The reduced scale model represents an idealised city district (Figure 1a), made up by parallelepipeds with squared section, with lateral size L= 250 mm and height H=50 mm high, representing 20 m high buildings at the 1:400 scale. Experiments were performed for different buildings configurations, obtained by varying the blocks spacing and the wind direction. A stationary source of a passive scalar Q was placed at a street intersection of coordinates X=Y=0 and Z=0.5H. Ethane was used as passive tracer and its concentration downwind the source was measured by means of a Flame Ionisation Detector. In the present study we focus a single configuration of obstacles, with regularly spaced buildings forming streets of width W= H and a wind direction of 12.5° (Figure 1a -dashed line). Concentration profiles measured within the streets for increasing distance from the source are shown in Figure 1a. The mean concentrations are expressed in a standard dimensionless form as , where C is the measured mean concentration in ppm,