Street canyon aerosol pollutant transport measurements

I.D. Longley
*
, M.W. Gallagher, J.R. Dorsey, M. Flynn, K.N. Bower, J.D. Allan
Physics Department, University of Manchester, Institute of Science and Technology (UMIST), P.O. Box 88, Manchester M60 1QD, UK
Accepted 1 April 2004
Abstract
Current understanding of dispersion in street canyons is largely derived from relatively simple dispersion models. Such
models are increasingly used in planning and regulation capacities but are based upon a limited understanding of the transport
of substances within a real canyon. In recent years, some efforts have been made to numerically model localised flow in
idealised canyons (e.g., J. Appl. Meteorol. 38 (1999) 157689) and stepped canyons (Assimakopoulos V. Numerical modelling
of dispersion of atmospheric pollution in and above urban canopies. PhD thesis, Imperial College, London, 2001) but field
studies in real canyons are rare.
To further such an understanding, a measurement campaign has been conducted in an asymmetric street canyon with busy
one-way traffic in central Manchester in northern England. The eddy correlation method was used to determine fluxes of size-
segregated accumulation mode aerosol. Measurements of aerosol at a static location were made concurrently with
measurements on a platform lift giving vertical profiles. Size-segregated measurements of ultrafine and coarse particle
concentrations were also made simultaneously at various heights. In addition, a small mobile system was used to make
measurements of turbulence at various pavement locations within the canyon.
From this data, various features of turbulent transport and dispersion in the canyon will be presented. The concentration and
the ventilation fluxes of vehicle-related aerosol pollutants from the canyon will be related to controlling factors. The results will
also be compared with citywide ventilation data from a separate measurement campaign conducted above the urban canopy.
D 2004 Elsevier B.V. All rights reserved.
Keywords: Street canyon; Dispersion; Urban aerosol; Traffic-induced turbulence
1. Introduction
Fine aerosol, when breathed in, has increasingly
been linked with a wide range of adverse health
effects (EPAQS, 2000, Harrison and Yin, 2000,
COMEAP, 2001). It is well established that one of
the major sources of fine aerosol is road vehicle
emissions and, hitherto, the principal strategy for
reducing the impact of fine aerosol on health has
been to reduce the emissions per vehicle, largely
based upon a mass metric, such as grams per kilo-
metre. However, health effects are presumably related
to personal exposure to ambient concentrations and
we do not expect emission cuts and reductions in
ambient concentrations to be linearly related. The
assessment, therefore, of such health impacts, and
any control strategy that may be implemented
depends upon a better understanding of how concen-
trations are controlled.
0048-9697/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.scitotenv.2004.04.072
* Corresponding author. Tel.: +44-161-2000-3961; fax: +44-
161-200-3951.
E-mail address: ian.longley@umist.ac.uk (I.D. Longley).
www.elsevier.com/locate/scitotenv
Science of the Total Environment 334335 (2004) 327336
More and more attention is being focussed upon
street canyons. These locations can contain a high
spatial concentration of sourcesvehiclesbut also
high concentrations of people, whether they be pedes-
trians, drivers, passengers or people who work or live
in the street and the buildings lining it. In such
canyons, dispersion is constrained by the buildings
and complicated by recirculation and the enhanced
and highly localised turbulence associated with such a
complex space (Vardoulakis et al., 2003). Some dis-
persion models used in Air Quality Management, such
as ADMS-Urban (McHugh et al., 1997) underestimate
the concentrations in street canyons (DEFRA, 2000).
The aerosol that is released from the canyon will be
transported across the city, and much will be exported
from the urban area. Once released from the urban
canopy layer, the aerosol can alter the chemical and
radiative properties of the atmosphere before being
deposited to potentially fragile ecosystems.
Air Quality Management, as well as other air
pollution studies, rely on dispersion modelling to link
emissions to concentrations; however, their weak-
nesses can be related to the simplified transport mech-
anisms they employ. Fine variations in space and time
are rarely modelled, if at all, although measurements
indicate that concentrations can vary hugely over a few
meters or a few minutes (or less). Some models do not
explicitly model street canyons at all. Some model
them, but take no account of asymmetry. The effects
deriving from the motion of the sources (momentum
transfer, extra turbulence, intermittency) are often not
modelled, nor well understood.
In particular, the different aerodynamic transport
behaviour of differently sized aerosol is not, as yet,
accounted for in regulatory dispersion models. This is
particularly important in light of the controversy
surrounding the issue of where the toxic fraction lies
within the spectrum of airborne particulate matter
(EPAQS, 2000; Harrison and Yin, 2000). Currently,
PM
10
is regulated and modelled; however, if we wish
to model fine and ultrafine aerosol separately, then we
would need a more sophisticated transport mechanism
to account for the way transport efficiency varies with
size and wind speed. Data from a previous project,
Fig. 1. Emission velocity (V
e
= F
N
/v
N
) as a function of wind speed (u) and particle diameter (D
p
). Data from SASUA project (Fowler et al.,
2002) derived with an ASASP-X and APS. The white colour indicates regions in which the net flux was deposition.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 328
Sources And Sinks of Urban Aerosol (SASUA),
described in Dorsey et al. (2002) and Fowler et al.
(2002), illustrates such variation of emission velocity
(defined as flux/concentration) in the urban canopy
(Fig. 1).
In order to investigate these issues, a pilot field
study was undertaken in an urban street canyon,
called Street Canyon Aerosol Research (SCAR).
The aim was to not just measure concentrations of
aerosol, but to attempt to understand the fine scale
motion of aerosol within a canyon. Eddy correlation
techniques had been successfully used higher in the
urban canopy to measure particle fluxes during
SASUA (see above) and so it was decided to extend
these techniques down into the street canyon. The
principal objective was to attempt to derive a param-
eterisation for observed aerosol emission flux as a
function of particle size.
2. Methodology
An experimental site was chosen alongside the
Town Hall building in Princess Street, Manchester,
one of the major cities in the northwest of England.
Manchester is prone to frequent cyclonic conditions
and experiences relatively high winds with a January
mean of 5 m s
1
and a July mean of 3.9 m s
1
. The
City of Manchester is surrounded by numerous satel-
lite towns making up the conurbation of Greater
Manchester, which has a population of 2.5 million.
The city centre has distinct peaks in weekday traffic
flow that occur at 79 and 1618 h local time.
The chosen section of the street (Figs. 2 and 3) is
120 m long and asymmetric with the Town Hall along
the southwest side, with a height of 2228 m, and a
variety of buildings on the northeast side of heights
1018 m. It has open squares at both ends (Albert
Square and St. Peters Square) and the canyon is
aligned at 130j to north. The street canyon itself is
17 m wide with two lanes of traffic both travelling
towards the southeast (the direction referred to as
down-canyon throughout SCAR), plus parking bays
on the south side and a row of bus stops on the north
side. Traffic is regulated by signals at both ends of the
canyon and traffic flow varied between 100 h
1
in the
early morning periods to 1200 h
1
during the evening
peak with speeds rarely exceeding 50 km h
1
.
Four experiments were performed. SCAR-1, -2 and
-3 in February, April and May 2002 lasted a week
each and concentrated on turbulence measurements
with sonic anemometers. SCAR-4 lasted for 2 weeks
in October 2001 (15th to 26th inclusive) and ran
continuously for each Monday to Friday period.
During SCAR-4, coarse, fine and ultrafine aerosol
instruments, as well as two ultrasonic anemometers
were placed upon a platform scissor lift (see instru-
ment details in Table 1), which was raised in 15-min
Fig. 2. The Princess Street experimental canyon site and wind sectors described in text.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 329
stages to give data at 2, 7, 12 and 15 m heights. A
mobile anemometer system was also deployed in three
of the experiments to investigate horizontal variation
of turbulence within the canyon. In each experiment,
the fixed instruments and the platform lift were
located in the parking bays on the southwest side of
the canyon (the higher wall), roughly at the midpoint
of the canyons length (see Fig. 3).
Four aerosol instruments were deployed, three on
the platform: the SMPS (TSI Model 3936 Scanning
Mobility Particle Sizer, covering the diameter range
4160 nm), ASASP-X (PMS Model Active Scatter-
ing Aerosol Spectrometer Probe, covering the diam-
eter range 0.13 Am) and the FSSP-100 (PMS Model
Forward Scattering Spectrometer Probe, diameter
range 230 Am). An RM Young Model 81000
ultrasonic anemometer was mounted in the centre
of the lift platform with its sensor head 3 m above the
platform floor. The inlet for the ASASP-X was
placed alongside the anemometer, whereas the SMPS
inlet was 1 m lower. The FSSPs sample head was
located 0.5 m above the platform floor on the
roadside edge of the platform. The anemometer,
ASASP-X and FSSP were logged together as an
eddy correlation system via a PC with a multiserial
card installed using software developed at UMIST.
The anemometer was logged at 20 Hz, and the
ASASP-X and FSSP at 1 Hz. The SMPS completed
a scan every 60 s. After quality control, these data
were averaged into 10-min periods.
A second ASASP-X was located in a trailer parked
behind the platform lift. It was logged as part of an
Table 1
Instruments deployed during SCAR referred to in the text
Instrument Measuring Location
R2 Solent anemometer 3D wind speed, temperature Mobile (pavement sites)
R2 Solent anemometer 3D wind speed, temperature Telescopic mast (215m)
RM Young anemometer 3D wind speed, temperature Scissor lift platform
SMPS (TSI 3936) Ultrafine aerosol number spectra,
10 nm< D
p
< 470 nm
Parking bay (SCAR-1)
scissor lift platform (SCAR-4)
PMS ASASP-X Fine aerosol number spectra and fluxes,
100 nm< D
p
<3 Am
Parking bay
PMS ASASP-X Fine aerosol number spectra and fluxes,
100 nm< D
p
<3 Am
Scissor lift platform
PMS FSSP-100 Coarse aerosol concentrations,
2 Am<D
p
<30 Am
Scissor lift platform
Fig. 3. Plan of experimental canyon site.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 330
eddy correlation system using a Gill Solent ultrasonic
anemometer. This was fixed on a mast so that the
sensor head was 3.5 m above the road, and was left in
a fixed position behind the trailer in a parking bay for
the duration of the experiment. This system was
logged at 20 Hz.
The platform lift was diesel powered. To avoid its
exhaust contaminating the experiment, the data
recorded during the 5 min starting with the diesel
engine being switched on was discarded. Inspection of
the data indicated that this period was sufficient to
allow for the dispersion of the extra aerosol. Other
data was discarded due to rain interfering with the
ultrasonic anemometers and periods of poor quality
power supply. After quality control, all data was
averaged over 10-min periods.
3. Results
3.1. Aerosol size distributions
Aerosol size distributions were derived by com-
bining data from the SMPS, ASASP-X and FSSP.
The mean aerosol number size distribution (Fig. 4)
observed over the 2-week period at the lowest mea-
surement height exhibited a clear mode at 2530 nm
in diameter. This same modal diameter was found
when the SMPS was used to sample the exhaust from
the UMIST Physics Department Land Rover Discov-
ery (turbocharged diesel engine, Fig. 4). In this case,
the vehicle was located within the UMIST grounds,
away from other aerosol sources, and the engine was
not under load. However, the same modal diameter
has also been observed in the urban canopy and has
been related to motor vehicle (especially petrol)
emissions (Williams et al., 2000). Thus, the observa-
tion of the 2530 nm mode in SCAR is consistent
with expectations for a receptor only a few meters
from busy traffic. The mass size distribution (Fig. 5)
was bimodal, with a mode around 100300 nm and
another around 10 Am.
3.2. Urban aerosol composition measurements
Time and size resolved chemical composition
measurements of aerosol were also sampled 3 months
later at a nearby street canyon site, located 25 m
above the street, close to the top of the canyon using
an Aerosol Mass Spectrometer (AMS). Full details are
given in Allan et al. (2003). The data clearly indicated
that the aerosol sampled in Manchester, in common
with other urban areas, had two mass modes. The
finer mode covered the diameter range 30300 nm
(therefore corresponding to the majority of the num-
ber of particles) and was largely composed of aliphat-
Fig. 4. Mean roadside aerosol size distributions at street-level (SCAR) and from UMIST Physics Department Land Rover exhaust (engine at
1000 and 3000 rpm).
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 331
ic organics representative of unburnt petrol and diesel
fuel. Mass concentrations of particles of diameter less
than 200 nm, which were identified as organic, were
strongly correlated with NO
x
concentrations that were
measured near the roof of the Town Hall. This
compares well with results obtained by Kleeman et
al. (2000), in which different engine types produced a
mode at 100200 nm, consisting mainly of organic
carbon (OC). These observations confirm that the
aerosol number concentrations are dominated by fine
particles that are mostly organic and derived from
vehicle emissions within the subject city. This also
agrees with the findings of Williams et al. (2000),
who found that Manchester ambient particle number
concentrations of the size range 100500 nm to be
linked to traffic activity. The second (coarser) mass
mode is dominated by sulphate and nitrate material,
which is generally advected into the canyon with
occasional contributions from oxygenated organic
compounds that have undergone significant atmo-
spheric processing. This data is described by Allan
et al. (2003) and is representative of several other
urban sites sampled.
3.3. Total aerosol number concentrations and mean
flow regimes
Total number concentrations of fine and ultrafine
aerosol (4.6 nm<D
p
< 160 nm) exhibited a diurnal
pattern that closely followed the traffic flow rate.
Daytime concentrations were generally four times
greater than those observed during the nighttime.
There were no discernable changes to the shape of
the particle size distributions. Variations in traffic flow
past the instruments, plus transient variations in air
flow (due to traffic induced turbulence at the 4-m
level) led to a large degree of variation in concen-
trations from minute to minute.
Data from the ultrasonic anemometers indicated
that the mean flow pattern within the canyon could be
divided into three regimes:

Up-canyon flow is defined as flow when the wind

direction above the canyon is within F40j of the
canyon axis, blowing against the traffic flow. In
such cases, the wind was channelled along the
canyon with very little vertical component.

NE perpendicular occurred when the wind

approached the canyon over the shorter NE wall.
In this case, a vortex flow was seen within the
canyon with exhaust pollutants blown away from
the instruments. There was only limited data in this
regime.

SW perpendicular occurred when the wind

approached over the taller wall. Although there
was some evidence of vortex flow it appeared that
this vortex might not have penetrated the full depth
of the canyon. Nevertheless, at street level the wind
was mostly blowing exhaust emissions towards the
instruments.
Fig. 5. Mean roadside aerosol volume size distribution at street level.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 332

Down-canyon winds did not occur during

SCAR-4.
Total number concentrations were found to be
dependent upon roof-level wind (Fig. 6). For up-
canyon and NE perpendicular flow, concentrations
increased as wind speed fell. For SW perpendicular
winds, number concentrations seemed unrelated to
wind speed; however, it was also found that wind
speed within the canyon was largely unrelated to
wind speed at roof level in this regime, suggesting
that air flow at street level became isolated from the
flow above in this regime. The mean number spectra
illustrate the effect that the regime had on roadside
aerosol populations (Fig. 7). Although the data for
the coarser aerosol is still being processed, early
indications are that larger particles were behaving
differently, with evidence of wind-driven resuspen-
sion increasing concentrations at higher wind
speeds.
Comparing data from the four heights revealed a
vertical concentration gradient. Total number concen-
trations at 17 m were generally f50% those at 4 m
during the day. The gradient was much reduced at
night. In SW perpendicular flow, the number spectrum
at 17 m was slightly broadened towards larger par-
ticles compared to 4 m (Fig. 8). This did not occur in
up-canyon flow. As the SW perpendicular regime
meant that the lower part of the canyon was relatively
isolated, it is possible that this might indicate ageing
Fig. 7. Ultrafine mean number size distributions at street level for the three principal wind directions (daytime at 4 m).
Fig. 6. Ultrafine number concentrations are controlled by wind speed and direction above the canyon (daytime at 4 m).
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 333
of the aerosol in situ, an idea that might be supported
by data from the AMS.
3.4. Aerosol emission fluxes
The average total number emission flux in the
range 100 nm< D
p
< 3 Am (Fig. 9) varied from around
100 cm
2
s
1
at night to 1000 cm
2
s
1
in the
middle of the day at 5 m. The flux was seen to be
related to both source strength (traffic flow) and
sensible surface heat flux. At 18 m, the flux followed
the same general pattern, but was more closely related
to heat flux than traffic flow. This result was consis-
tent with that found in the urban canopy layer in the
SASUA project in Edinburgh (Dorsey et al., 2002).
The contribution of wind-dependent resuspension of
larger particles, thermal stability and the different
behaviour of different-sized subranges is under inves-
tigation for future publication.
3.5. Turbulence
Turbulence data was combined from all four
experiments. Due to the prevalence of positive
stresses and the dependence of stress upon wind
direction, it was found that the variances were best
Fig. 9. Diurnal average of fine mode (100 nm< D
p
< 3 Am) vertical emission number flux at lowest and highest measurement levels.
Fig. 8. Ultrafine mean number size distributions at lowest and highest measurement levels in SW perpendicular winds only.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 334
presented as standard deviations normalised by local
sonic wind speed, i.e, r
u,v,w
/U. r
w
/U was well related
to sonic wind speed U (Fig. 10), with increased values
when U< 1.5 m s
1
, indicating the relative impor-
tance of other sources, such as thermal and traffic-
induced turbulence, when wind speeds are low.
If it is assumed that r
w
is made up of a contribution
from the mean wind and a contribution from the
traffic (r
wt
), then it can be suggested that r
w

aU
2
r
2
wt

1=2
.
From Solent data, a was found to be 0.16 at a
height of 3.5 m. It is possible that a rises with
heightthis is still being investigated. r
wt
was found
to follow the relationship r
wt
= 0.0135 T
1/2
at 3.5 m,
where T is traffic flow rate in vehicles per hour. The
correlation between r
wt
and traffic becomes poorer
with increased height.
A vertical profile of r
w
/U (when U>1.5 m s
1
,
Fig. 11) shows a weak positive gradient at night. r
v
/U
followed a similar pattern to r
w
/U, but r
u
/U was
different in having larger values.
The mobile turbulence system revealed that hot
spots of r
w
/U occurred wherever the ratio of sonic
wind speed in the canyon to wind speed measured at
roof level, U/U
roof
fell below about 0.3. This was more
likely to happen in the sheltered zone at the down-
Fig. 10. Dependence of normalised vertical turbulence r
w
/U upon wind speed and time of day, including parameterisation, at 3.5 m height.
Fig. 11. Vertical profiles of normalised turbulence when U>1.5 m s
1
.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 335
wind end of the canyon, which was larger in more
perpendicular winds, and when the wind blew over the
higher canyon wall. Hot spots also occurred where
winds blowing from opposite ends of the canyon met.
4. Conclusion

The flux of accumulation mode aerosol (100

nm< D
p
<3 Am) has been measured in a city
centre street canyon and found to be related to
surface sensible heat flux, traffic activity, wind
speed and direction.

Street-level aerosol concentrations have been

shown to be influenced by asymmetrical vortex
flow within the canyon.

A vertical gradient was found in fine aerosol

number concentrations below roof level.

A weak positive vertical gradient was seen in

turbulent variances.

A parameterisation has been derived for r

w
relating
it to local wind speed and direction and traffic flow
rate.
5. Forthcoming analysis
A flux budget will be attempted to fully describe
the movement of particles within the canyon and
relate the balance between vertical and horizontal
fluxes to controlling factors. Further data analysis will
allow a fuller description of the behaviour of different
sizes of aerosol. Work will proceed to achieve an
improved parameterisation for wind-driven and traf-
fic-induced turbulent resuspension. This will be aided
by data from Salfex, measuring turbulence in a
suburban street canyon.
Acknowledgements
This work was supported by CERC Ltd, Cam-
bridge, UK.
References
Allan JD, Alfarra MR, Bower KN, Williams PI, Gallagher MW,
Jimenez JL, et al. Quantitative sampling using an Aerodyne
aerosol mass spectrometer: Part 2. Measurements of fine partic-
ulate chemical composition in two UK cities. J Geophys Res,
Atmos 2003;108(D3):4091 (doi:10.1029/2002TD002359).
Committee on the Medical Effects of Air Pollutants (COMEAP).
Statement on long-term effects of particles on mortality. London:
HMSO;, 2001.
Department for Environment, Food and Rural Affairs (DEFRA).
First phase air quality review and assessment. A summary.
London: HMSO; 2000.
Dorsey JR, Nemitz E, Gallagher MW, Fowler D, Williams PI,
Bower KN, et al. Direct measurements and parameterisation
of aerosol flux, concentration and emission velocity above a
city. Atmos Environ 2002;36:791800.
Expert Panel on Air Quality Standards (EPAQS). Airborne particles:
what is the appropriate measurement on which to base a stan-
dard? London: HMSO;, 2000.
Harrison RM, Yin J. Particulate matter in the atmosphere: which
particle properties are important for its effects on health? Sci
Total Environ 2000;249:85101.
Kleeman MJ, Schauer JJ, Cass GR. Size and composition distribu-
tion of fine particulate matter emitted from motor vehicles. En-
viron Sci Technol 2000;34:113241.
McHugh CA, Carruthers DJ, Edmunds HA. ADMS-Urban: an air
quality management system for traffic, domestic and industrial
pollution. Int J Environ Pollut 1997;8:66674.
Vardoulakis S, Fisher BEA, Percleous K, Gonzales-Flesca N. Mod-
elling air quality in street canyons: a review. Atmos Environ
2003;37:15582.
Williams PI, Gallagher MW, Choularton TW, Coe H, Bower KN,
McFiggans G. Aerosol development and interaction in an urban
plume. Aerosol Sci Technol 2000;32:1206.
I.D. Longley et al. / Science of the Total Environment 334335 (2004) 327336 336