Environmental Engineering Reference
In-Depth Information
coherent structures and is able to reproduce prescribed first and secondary order
one point statistics and turbulence length-scales. The overall performance of the
model used to simulate turbulent wall bounded flows has been evaluated using a
channel flow numerical simulation and a flow past a bluff body physical simulation
(see Pavlidis et al., 2009). Transport of pollutant concentrations is determined by a
high resolution method, which is globally high order accurate in space and time.
3. Results
The area of study is the intersection of Marylebone Rd. and Gloucester Place, in
Central London, UK. A large number of field and wind tunnel experiments (scale
1/200) have been undertaken for this site for the DAPPLE project (Arnold et al.,
2004). The area of interest is 800 × 600 m² and includes 48 buildings. Available
data include mean and turbulent quantities of flow and concentration from a number
of tracer releases for a number of receptors. At low ambient wind speeds (~3 m/s
at average roof-top height) the wind tunnel data deviate from the field data. In this
study this low ambient wind speed scenario is reproduced numerically. The
simulation models the dispersion of a 15 min release of a passive tracer from a
fixed source for 30 min.
A schematic of the area under study along with the mean flow direction, the
tracer source and receptors locations considered are given in
Fig. 1.
The lateral
and vertical extents of the computational domain have been kept identical to those
of the wind tunnel leading to a 200 m deep boundary layer. An anisotropic adaptive
mesh with a maximum of 750,000 nodes is used. The mesh is adapted every 15
time-steps. For the generation of realistic inflow boundary conditions the synthetic
eddy method (Pavlidis et al., 2009) is employed, the inflow plane is placed two
boundary layer depths upwind of the built-up area to allow the flow to develop.
Data used to prescribe the inlet boundary conditions come from the wind tunnel.
No-slip boundary conditions are applied on all solid walls. An adaptive time-
stepping scheme is employed to satisfy a CFL number equal to four. A snapshot of
the surface mesh and tracer concentration at the end of the simulation (30 min) is
shown in
Fig. 3.
Comparison of total normalized (with velocity at roof-top height,
U
, roof-top
height,
H
, and emission rate,
Q
) concentration, at the sampling points between
field, two (identical) physical model simulations and Fluidity is given in
Fig. 2.
Note that there are no data for receptor R2 for the field and receptor R3 for
Fluidity. It is evident that the wind tunnel tends to over-predict the concentrations,
while Fluidity is closer to the field data.
In particular, the solution obtained from the receptor at roof-top height (R5)
agrees with field data, while one of the wind tunnel simulations is very close as
