Tracer dispersion experiments were performed in an atmospheric boundary layer wind tunnel. An urban center of uniform city blocks and building heights was simulated at a 300:1 scale; street canyon aspect ratio was 1.2:1. Ethane (C2H6) was released at a constant rate from a dual line source located in the street canyon for three experiments and in the orthogonal 'avenue canyon' for one series of experiments. Receptor arrays were located in the street canyon; 15-28 receptors were positioned on each of the upwind and downwind building faces in rectangular arrays. Wind direction was varied over 90-degrees in 10-degrees increments. Four dispersion experiments were undertaken and included both rectangular and square city blocks as well as variable avenue widths; one series placed the emission source in the orthogonal avenue canyon. Several significant sensitivities of street-canyon concentrations to the shape of the city block and to emissions from adjacent, orthogonal streets (i.e. avenues) were observed. Mid-block concentrations for the rectangular block are a factor of two greater than the square block (avenues wider than streets in both cases); rectangular blocks have the concentration maximum in mid block whereas square blocks have maxima near the ends. When the avenue width is reduced to equal the street width (square blocks), concentrations increase about 25 per cent on average. Entrainment from a single adjacent avenue yields concentrations in the street that are a significant fraction of those from an equivalent source in the street. These observations have major implications for current street-canyon modeling practices. A new modeling approach is introduced to simulate the shape of the upwind and downwind concentration profiles and their sensitivity to wind direction. The puff-based model considers two diffusion regimes: vortex and purging. Preliminary model results are very encouraging when compared to wind tunnel observations for a rectangular block and wind angles of 0-60-degrees.
