This paper investigates numerical method for predicting wind field and pollutant dispersion in an urban area with crowded buildings and heavy traffic. Large eddy simulation is employed in numerical computation with Lagrangian dynamic model to close the subgrid scale stress. The fine grids are required in numerical simulation of micro-atmospheric flows in an urban area with crowded and complex building-streets. To realize the numerical simulation on available computer resources a domain decomposition method and a combined model for the building-street layout are proposed that the computed domain of an urban area is decomposed into a central area, where the pollution is the major concern and a surrounding region where the pollutant distribution is not concerned. The combined flow model is proposed for the complex building-street layout that the fine grid mesh and exact flow boundary conditions are prescribed in the central area while the buildings are treated as roughness elements in the surrounding region where the coarse grids are used. The proposed numerical method is applied to simulate the wind field and pollutant dispersion from vehicle exhaust in the Rua Do Campo area of Macau (referred to the site hereafter). The wind speed and temperature fields are computed from 7 am on September 6 2005 to 7 am of the next day and the pollutant concentration field is computed from 7 am to 6 pm on September 6 2005. The results show fairly good agreement with the field observation. The wind field and pollutant concentration distribution are demonstrated in the site by proposed combined model and they are compared with those computed by drag element model in whole computational domain. The results show that the combined model is an appropriate method for predicting wind field and pollutant dispersion in the crowded building area. The paper also reveals the relevance of buoyancy effect in the prediction of urban atmospheric environment. © 2011 Elsevier Ltd.