A simple urban dispersion model is tested that is based on the Gaussian plume model and Briggs' urban dispersion curves. A key aspect of the model is that an initial dispersion coefficient (σ) of 40 m is assumed to apply in the x, y, and z directions in built-up downtown areas. This initial σ accounts for mixing in the local street canyon and/or building wakes. Observations from tracer experiments during the Madison Square Garden 2005 (MSG05) field study are used for model testing. MSG05 took place in a 1 km by 1 km area in Manhattan surrounding Madison Square Garden. Five different perfluorocarbon tracer (PFT) gases were released concurrently from five different locations around MSG, and were sampled by 20 samplers near the surface and seven samplers on building tops. There were two separate continuous 60 minute tracer release periods on each day, beginning at 9 am and at 11:30 am. Releases took place on two separate days (March 10 and 14). The samplers provided 30 minute averaged PFT concentrations from 9 am through 2 pm.

This analysis focuses on the maximum 60-minute PFT gas concentration at each sampler location for each PFT for each release period. Stability was assumed to be nearly neutral, because of the moderate winds and the mechanical mixing generated by the buildings. Input wind direction was the average observed building-top wind direction (285° on March 10 and 315° on March 14). Input wind speed was the average street-level observed scalar wind speed (2 m/s for both days). To be considered in the evaluation, both the observed and predicted concentration had to exceed the threshold.

Concentrations normalized by source release rate, C/Q, were tested. The data for releases around MSG were separated into surface samplers and building top samplers. The surface samplers were further separated into those at distances less and greater than 100 m. For the larger distances > 100 m, containing most of the data, the median observed and predicted C/Q for surface samplers are within 4% of each other, and 52% of the predictions are within a factor of two of the observations. For the building top samplers, the median predicted concentration is about 60% of the median observed concentration. For the one sampler at a downwind distance less than 100 m, the model underpredicted by a factor of four, suggesting that at short distances (i.e., when the release is in the same street canyon as the receptor and there are no obstructions in between), the initial σ should be reduced to about 10 m.