1. Large-Eddy Simulation of plume dispersion within various actual urban areas
Hiromasa Nakayama*, Klara Jurcakova** and Haruyasu Nagai*
*Japan Atomic Energy Agency, Japan
**Academy of Sciences of the Czech Republic, Prague

2. Local-scale atmospheric dispersion models for emergency response
Atmospheric dispersion problems within urban areas
Accidental release of chemical materials
Intentional release of hazardous materials by terrorist attack
Prediction of spatial extent of contaminated regions using building-resolving CFD models
M2UE (Danish Meteorological Institute)
FAST3D-CT (U.S. Naval Research Laboratory)
FEM3MP (U.S. National Atmospheric Release Advisory Center)
MSS (French Atomic Energy Commission)
LOHDIM-LES (Japan Atomic Energy Agency)
In local-scale atmospheric dispersion problems, we have important issues such as accidental release and intentional release of hazardous materials within urban area. For investigating spatial extent of contaminated areas, it is well known that numerical simulation technique using CFD is very useful. Many researchers have developed emergency response tool. Our group team also has developed LES-based Local-scale High-resolution Dispersion model. These urban dispersion models resolve individual urban buildings and can provide detailed information on turbulent flow patterns and spatial distribution patterns of plume concentrations.
These urban dispersion CFD models can provide detailed information on turbulent flow patterns and spatial distributions of plume concentration within urban areas.

3. What is the spatial extent of distribution patterns of concentration influenced by urban surface geometry?
Schematic of wind flow in urban area
(Ratti et al and Nakayama et al. 2012)
Urban areas consisting of buildings with variable height and obstacle density
Strong three-dimensionality of the turbulent flows within urban canopy
Complex spatial distribution patterns of plume concentration
In urban areas, the lower part of boundary layer is called as roughness sub-layer and especially in urban canopy layer the strong three-dimensionality of turbulent flows are formed by the influence of each urban obstacle. As a result, distribution patterns of plume concentrations become highly complex.
Therefore, in numerically simulating plume dispersion, an important issue remains in setting up enough computational domain size in order to capture distribution patterns of plume concentration influenced by urban surface geometries. This table shows roughness parameters of European, North American, and Central Tokyo. Building height variability is defined as the ratio of the standard deviation of building height to the average building height. Roughness density is defined as the ratio of the total frontal area of roughness elements to the total surface area.
Building height variability values of North America and Central Tokyo surface geometries of North America and Central Tokyo are found to be highly variable and show much larger than the ones of Europe. Roughness density values of Salt Lake City and Berlin are found to be smaller than the other areas. From these data, urban surface geometries show a wide variety due to a wide range of building height variability and roughness density.
In this study, our objective is to carry out a series of LESs of plume dispersion in various urban areas and clarify the spatial extent of the distribution patterns of influenced by urban surface geometry by comparative analysis.
The extent to which a specific air dispersion model is suitable for the evaluation of air toxic source impacts depends upon several factors
Objective
Carry out a series of LESs of plume dispersion in urban areas with a wide range of building height variability and obstacle density
Clarify the distribution patterns of plume concentration influenced by urban surface geometry by comparative analysis

4. Model structure of LOHDIM-LES developed at JAEA
Recycling only fluctuating components of wind velocity
Turbulent Flow
Turbulent Inflow
Fully-developed urban boundary layer flow
Spatially-developing rough-wall boundary layer flow
Tripping Fence
Roughness Blocks
Urban area
Driver region
Main analysis region
Schematic diagram
Procedure of calculating
Generate basic boundary layer flow by recycling method(Kataoka&Mizuno,2002)
Produce active turbulent fluctuations by various roughness obstacles
Carry out LESs of turbulent flow and plume dispersion in urban areas
Basic equations
Spatially-filtered continuity equation, Navier-Stokes equation, and Scalar conservation equation
SGS turbulent and scalar effects
For flow field, the standard Smagorinsky model(1963) with the constant of 0.1
For dispersion filed, the standard Smagorinsky model with the turbulent Schmidt number of 0.5
Building effects
Immersed boundary method by Goldstein et al. (1998)
This figure shows a model structure of LOHDIM-LES developed at JAES. We set up a driver region for generating spatially-developing boundary layer flow and main analysis region. First, by recycling method, basic turbulent boundary layer flow is generated and active turbulent fluctuations are produced by various roughness obstacles. Then, by imposing this turbulent inflow data at the inlet boundary of main region, LES of plume dispersion in urban area is carried out. 4

5. Resolved actual urban area
Computational conditions
Computational model
Driver region
Main analysis region
Resolved actual urban area
Mesh number
460×250×90
475×250×90
375×250×90
Domain size
5.5km×1km×1km
2.0km×1km×1km
1.5km×1km×1km
Grid resolution
4.0m-20.0m×4.0m×
1.3m-53m
4.0m×4.0m×
Boundary
Flow field
Dispersion field
Inlet
Driver region
Recycling technique (Kataoka & Mizuno, 2002)
Main region
Turbulent inflow generated in the driver region
Exit
Conventional convective type
Top
Ground
Side
Periodic
Building effects
External force
(Goldstein et al.,1993)

6. Cases of comparative analysis for
wind tunnel experiments（Bezpalcova,2008）and LESs
Surface geometry type
Average building height[m]
Building height variability[-]
Obstacle density[-]
Idealized urban canopy
(Bezpalcova,2008)
Cubic buildings array
42.0
0.0
0.16
0.25
0.33
Actual urban site in Central Tokyo (present LESs)
Low-rise buildings area
9.6
0.60
0.39
Street-canyon area
22.7
0.71
0.56
Complex of high-rise and low-rise buildings area
26.5
0.85
0.52
High-rise buildings area
34.1
0.94
In this study, we set up two types of urban surface geomtries.

7. Building heights distributions of Central Tokyo used in LESs
100m 0m
NNW
(a)Low-rise buildings area
(b)Street-canyon area
100m 0m
NNW
(c)Complex of high-rise and low-rise buildings area
(d)High-rise buildings area

8. Characteristics of approach flow generated in driver region
1km
1km
Tripping fence
Roughness blocks
0.5km
5.0km
These figure compares the LES results of turbulence characteristics of approach flow with the wind tunnel experimental data of Bezpalcova and Ohba (2008) and the recommended data of Engineering Science Data Unit (ESDU). ESDU provides comprehensive turbulence characteristics of neutrally stratified atmospheric boundary layer ranging from the ground surface to 300m. It is found that mean wind velocity of LES is consistent with experimental profile of 0.25 power law. Turbulence intensities for each component of the experiment lie within the ESDU recommended data of moderate rough and rough surfaces. Although w-component of the LES is consistent with the experimental data, u-component is underestimated and v-component is overestimated. But, each component of the LES also lie within the ESDU recommended data.
Mean velocity
Turbulence intensity for u-component
Turbulence intensity for v-component
Turbulence intensity for w-component

9. Spatial distributions of mean concentration at ground level
Source location
Source location
10-1
10-6
Cmean/Cinit
NNW
(a)Low-rise buildings area
(b)Street-canyon area
Source location
Source location
10-1
10-6
Cmean/Cinit
NNW
(c)Complex of high-rise and low-rise buildings area
(d)High-rise buildings area

10. Spatial distributions of r.m.s. concentration at ground level
Source location
Source location
10-1
10-6
Cr.m.s./Cinit
NNW
(a)Low-rise buildings area
(b)Street-canyon area
Source location
Source location
10-1
10-6
Cr.m.s./Cinit
NNW
(c)Complex of high-rise and low-rise buildings area
(d)High-rise buildings area

11. Mean concentration distributions along wind direction
from the point source

12. R.m.s. concentration distributions along wind direction
from the point source

13. Conclusion
We investigated spatial extent of the distribution patterns of plume concentrations within urban canopy.
Mean concentration distributions are nearly the same among various urban areas at a downwind distance of the point source greater than 1.0km.
The difference of r.m.s. concentration distributions among various urban areas also become small at a downwind distance of the point source greater than 1.0km.
In order to capture distribution patterns of plume concentrations within urban canopy, computational domain size should be at least 1.0km along wind direction from the point source.