Environmental Engineering Reference
In-Depth Information
1. Introduction
Atmospheric transport and dispersion models are used to predict downwind
hazards associated with the release of hazardous materials. In urban areas, con-
centration patterns may be spread out and distorted because wind flows are variable
due to vortices and channeling forced by the buildings. Three-dimensional wind
fields accounting for urban effects can be constructed through a variety of methods
ranging from simple urban canopy parameterizations to complex computational
fluid dynamics (CFD) models. In an attempt to strike a balance between model
speed and complexity, several groups have developed urban diagnostic wind flow
models based on the Röckle (1990) approach, which initializes the system by
parameterizing the geometric shapes of recirculating wakes and their flow patterns
behind, over, and around buildings. The approach requires a detailed represen-
tation of the 3-D building geometry (with resolution of about 1-5 m). Kaplan and
Dinar (1996) were the first to implement this method in an operational urban wind
flow and dispersion model. Subsequently, several other groups have modified the
basic model and coupled it with mass consistency algorithms to generate detailed
flow fields within the urban environment within a matter of minutes. Dispersion
models (usually Lagrangian particle dispersion models, or LPDMs) can then be
used to quickly predict concentrations of hazardous substances.
In this paper, simulations from four such models are compared using a subset
of the observations from the Joint Urban 2003 (JU2003) field study in Oklahoma
City, OK, USA (see Allwine et al., 2004; Clawson et al., 2005). Data from con-
tinuous release 1 during IOP 2 (day) and continuous release 2 during IOP 8 (night)
are used, and the release location was near the downtown Westin hotel.
The four models are the Los Alamos National Laboratory's (LANL) QUIC
model suite (Williams et al., 2004; Gowardhan et al., 2008; Singh et al., 2008), the
Science Applications International Corporation (SAIC)-ARIA MicroSwift/Spray
(MSS) model (Moussafir et al., 2004); the US Army Research Laboratory's (ARL)
three-dimensional Wind Field (3DWF) model (Wang et al., 2005), and the Israel
Institute of Biological Research (IIBR) Kaplan and Dinar (1996) model. All models
implement Röckle's original 1990 methodology but have been modified based on
observations from a number of recent urban tracer studies and wind/water tunnel
experiments. Because the Rockle models produce only simulations of the wind
field and not the turbulence, it is necessary to parameterize the turbulent speeds
and Lagrangian time scales for use in the LPDM.
2. Meteorological Inputs
The wind data used for model inputs in this study are 15-min average vertical
wind profiles collected with the Pacific Northwest National Laboratory (PNNL)
SODAR, located in a suburban area approximately 3 km S/SW of downtown
