Reliable transport and dispersion models are needed for making emergency response decisions in the event of a terrorist release of chemical or biological agents at street level in a built-up downtown urban area. The wind flows as well as the characteristics of turbulent dispersion must be known. Several recent field experiments have been carried out in order to improve urban modeling capabilities and demonstrate the accuracy of their predictions. This paper describes the objectives of the studies and the observations made during some of these experiments, including Urban 2000 (Salt Lake City), Los Angeles 2001, Barrio Logan 2001 (San Diego), and Joint Urban 2003 (Oklahoma City). In all cases, detailed meteorological and SF6 tracer gas observations have been made. The current paper surveys these field experiments and describes some procedures used for analysis of the data and for modification and evaluation of models.

At heights less than the building tops, the urban area is characterized by relatively low wind speeds, high turbulence intensities, and a tendency towards neutral stability. These effects are due to the large drag on the atmosphere, and the relatively high rates of generation of mechanical turbulence by the building obstacles. Methods of estimating wind and turbulence profiles in urban areas are summarized and are compared with the field data. Some basic characteristics of the SF6 tracer gas releases and the monitoring data are also summarized for the field experiments and general relations presented. A consistent finding is that there is greatly enhanced turbulent dispersion in all directions.

As an example of the use of the field data to evaluate transport and dispersion models, the HPAC urban suite of models is evaluated using the tracer data from the Urban 2000 experiment. The experiment consisted of six nights during which three one-hour releases of SF6 tracer were made at ground level in the downtown area. 30-minute averaged concentrations were observed at about 70 monitors located on arcs at distances from about 150 m to 6000 m downwind. Meteorological variables were observed at over 20 sites throughout the domain, including several vertical soundings. Five alternate assumptions for wind inputs were tested. For the various HPAC model combinations, the typical relative mean bias of 30 minute averaged concentrations was an overprediction by about 70 %, with a range from no mean bias to a factor of four overprediction. In general, the predicted and observed plumes were lined up satisfactorily in wind direction (within plus or minus 20 degrees). Several individual algorithms in the model were also evaluated, such as the modules for lateral and vertical dispersion and for effective cloud transport speed. The performance of the HPAC urban modules is shown to be within the range of satisfactory model performance for other models at other sites. However, the results are being further analyzed to better understand the relative performance of the modules with various wind inputs.