validation has been for wind tunnel tests and
engineering flows, where the kind of very-largescale eddies and downdrafts found in atmospheric
flows are absent. Such motions can lead to more
rapid exchange between the upper flow above and
within the canopy layer, as discussed in the review
by Mestayer and Anquetin [5].
As shown in Table 4, and reviewed by Britter and
Hanna [11], most FAMs for the neighborhood scale
tend to focus on equilibrium flows within and just
above the canopy. In a “porous” canopy (Fig. 2a),
the mean velocity within the canopy U c (z) is driven
by the turbulent shear stresses generated in the
intense shear layer just above the canopy. Here the
ratio of U c /U (z * ), where U(z * ) is the velocity at the
top of, or above, the roughness layer (Fig. 2),
depends on the porosity – a typical value being
0.3. In a non-porous canopy, typically in the inner
city, where flows are determined by canyon and
downdraft effects, this mean velocity ratio varies
considerably over the range 0.1–0.3 [23].
The approximate average velocity profile
(Eq. 2.1) is not applicable at the edges of “neighborhoods” where the density and heights of buildings vary, or in the interiors of such regions, such
as near parks, squares, etc. At the edges of porous
canopies, where the air flow in the canopy varies
rapidly, an FAM approximate model has been
developed based on the same approach as FCM
by representing the canopy as a porous layer and
(as for the FAM in the mesoscale range) solving
perturbation equations semianalytically to provide fast computation of the mean velocity and
shear stress [9, 56]. These have been verified
against field and wind tunnel studies.
Over non-porous canopies (typically categories (i)–(v) in Table 3), the determining parameters
(H c , z 0 (x, y)) are the same as for FCM over these
elevation/roughness changes. The linearized
FAM approaches give very similar results to
those using FCMs [57, 58].
Modelling the Building and Street Scale
The purpose of detailed flow modelling in and
around individual buildings and the surrounding
streets is firstly to understand how rapidly
released matter disperses locally. Secondly,
detailed studies of these local scale flows are
necessary, as explained in the previous section,
to develop models over neighborhood scales for
flow and dispersion.
Fully computational methods, using a variety
of turbulence modelling methods, have been
extensively applied to computing flows around
single buildings in turbulent boundary layers.
Using turbulence closure methods, many models
have predicted mean velocities near the buildings
within one or two widths; but such models (e.g.,
k-ε models) have tended to over-predict smallscale turbulence around the structure [59] and
under-predict the large-scale atmospheric eddies.
This leads to an over/under-prediction of the mean
velocity defects in the wakes downwind of 3-D/2-D
structures in the atmospheric boundary layer.
Only the computationally intense method of
Large Eddy Simulation (LES), or unsteady
modelling of the fluctuating flows using turbulence closure models, can accurately represent
the distortion of the large-scale eddy motion
around the building and its effect on downwind
wakes [60, 61].
Calculations using FCM with turbulence closure have also been performed on flow around
small groups of buildings using resolutions
down to about 0.1 H. The predictions of the turbulence in these flows are more reliable than those
of isolated structures because most of the turbulence in this instance is generated locally between
the buildings and is of smaller scale (similarly,
wind tunnel models of dispersion around buildings are generally more reliable when the buildings are in groups than when isolated because
wind tunnels cannot simulate the low frequency
fluctuation in wind direction). This is because the
primary interactions between the wake of one
building and its impact on adjacent buildings are
not sensitive to the structure of turbulence; however, the wakes of isolated or well-separated
buildings are strongly affected by the form of the
low frequency spectrum of turbulence. This is
why in some institutions, for example, the US
Naval Laboratory [62] and more recently at Los
Alamos [63], work is being undertaken on dispersion in dense urban areas with fast FCMs in which
approximate velocity fields are computed using
classical grid box methods of finite size (say 1 m)
but neglecting completely any modelling of turbulent stress. This reduces the number and the
178
Urban Air Quality: Meteorological Processes
engineering flows, where the kind of very-largescale eddies and downdrafts found in atmospheric
flows are absent. Such motions can lead to more
rapid exchange between the upper flow above and
within the canopy layer, as discussed in the review
by Mestayer and Anquetin [5].
As shown in Table 4, and reviewed by Britter and
Hanna [11], most FAMs for the neighborhood scale
tend to focus on equilibrium flows within and just
above the canopy. In a “porous” canopy (Fig. 2a),
the mean velocity within the canopy U c (z) is driven
by the turbulent shear stresses generated in the
intense shear layer just above the canopy. Here the
ratio of U c /U (z * ), where U(z * ) is the velocity at the
top of, or above, the roughness layer (Fig. 2),
depends on the porosity – a typical value being
0.3. In a non-porous canopy, typically in the inner
city, where flows are determined by canyon and
downdraft effects, this mean velocity ratio varies
considerably over the range 0.1–0.3 [23].
The approximate average velocity profile
(Eq. 2.1) is not applicable at the edges of “neighborhoods” where the density and heights of buildings vary, or in the interiors of such regions, such
as near parks, squares, etc. At the edges of porous
canopies, where the air flow in the canopy varies
rapidly, an FAM approximate model has been
developed based on the same approach as FCM
by representing the canopy as a porous layer and
(as for the FAM in the mesoscale range) solving
perturbation equations semianalytically to provide fast computation of the mean velocity and
shear stress [9, 56]. These have been verified
against field and wind tunnel studies.
Over non-porous canopies (typically categories (i)–(v) in Table 3), the determining parameters
(H c , z 0 (x, y)) are the same as for FCM over these
elevation/roughness changes. The linearized
FAM approaches give very similar results to
those using FCMs [57, 58].
Modelling the Building and Street Scale
The purpose of detailed flow modelling in and
around individual buildings and the surrounding
streets is firstly to understand how rapidly
released matter disperses locally. Secondly,
detailed studies of these local scale flows are
necessary, as explained in the previous section,
to develop models over neighborhood scales for
flow and dispersion.
Fully computational methods, using a variety
of turbulence modelling methods, have been
extensively applied to computing flows around
single buildings in turbulent boundary layers.
Using turbulence closure methods, many models
have predicted mean velocities near the buildings
within one or two widths; but such models (e.g.,
k-ε models) have tended to over-predict smallscale turbulence around the structure [59] and
under-predict the large-scale atmospheric eddies.
This leads to an over/under-prediction of the mean
velocity defects in the wakes downwind of 3-D/2-D
structures in the atmospheric boundary layer.
Only the computationally intense method of
Large Eddy Simulation (LES), or unsteady
modelling of the fluctuating flows using turbulence closure models, can accurately represent
the distortion of the large-scale eddy motion
around the building and its effect on downwind
wakes [60, 61].
Calculations using FCM with turbulence closure have also been performed on flow around
small groups of buildings using resolutions
down to about 0.1 H. The predictions of the turbulence in these flows are more reliable than those
of isolated structures because most of the turbulence in this instance is generated locally between
the buildings and is of smaller scale (similarly,
wind tunnel models of dispersion around buildings are generally more reliable when the buildings are in groups than when isolated because
wind tunnels cannot simulate the low frequency
fluctuation in wind direction). This is because the
primary interactions between the wake of one
building and its impact on adjacent buildings are
not sensitive to the structure of turbulence; however, the wakes of isolated or well-separated
buildings are strongly affected by the form of the
low frequency spectrum of turbulence. This is
why in some institutions, for example, the US
Naval Laboratory [62] and more recently at Los
Alamos [63], work is being undertaken on dispersion in dense urban areas with fast FCMs in which
approximate velocity fields are computed using
classical grid box methods of finite size (say 1 m)
but neglecting completely any modelling of turbulent stress. This reduces the number and the
178
Urban Air Quality: Meteorological Processes