Particle number size distributions have been measured simultaneously by
scanning mobility particle sizers (SMPSs) at five sites in central London for
a 1 month campaign in January–February 2017. These measurements were
accompanied by condensation particle counters (CPCs) to measure total particle
number count at four of the sites and Aethalometers measuring black carbon
(BC) at five sites. The spatial distribution and inter-relationships of the
particle size distribution and SMPS total number counts with CPC total number
counts and black carbon measurements have been analysed in detail as well as
variations in the size distributions. One site (Marylebone Road) was in a
street canyon with heavy traffic, one site (Westminster University) was on a
rooftop adjacent to the Marylebone Road sampler, and a further sampler was
located at Regent's University within a major park to the north of Marylebone
Road. A fourth sampler was located nearby at 160 m above ground level on the
BT tower and a fifth sampler was located 4 km to the west of the main
sampling region at North Kensington. Consistent with earlier studies it was
found that the mode in the size distribution had shifted to smaller sizes at
the Regent's University (park) site, the mean particle shrinkage rate being
0.04 nm s
The adverse health consequences of air polluted by particulate matter are
now well recognised (WHO, 2006). While the main focus has been on the public
health impact of exposure to fine particulate matter measured by mass
(PM
In addition to concerns over human health, there are other reasons for the study of the size distribution of airborne particles. Not only does this strongly influence their location and efficiency of deposition in the human lung, the particle size distribution can also be a strong indicator of particle source, with there being some clear differences between the modal diameter of particles arising from different sources (Vu et al., 2015a). The clearest distinction is between particles arising from combustion and other high-temperature sources, which tend to be predominantly very small, and particles generated by attrition processes which are typically far more coarse. However, even within the particles generated from combustion and other high temperature sources, there may well be different modal diameters associated with different sources or even multiple modes associated with an individual source (Vu et al., 2015a). For example, exhaust emissions from diesel engines typically comprise both a nucleation mode and an overlapping Aitken mode, reflecting in the former case particles comprised mainly of condensed lubricating oil formed after the combustion process, and in the latter case solid carbonaceous particles formed within the combustion process (Shi and Harrison, 1999; Alam et al., 2016).
After their emission, particle size distributions are also liable to change through dynamic processes. These include evaporation, which causes particles to shrink without changing the overall number; condensation, which causes particles to grow without a change in total number; coagulation, which also causes growth but reduces the total particle number; and deposition, which causes a reduction in number and is a strong function of the particle size.
There are detailed assessments of the concentrations and size distributions of nanoparticles in the rural atmosphere (Van Dingenen et al., 2004; Asmi et al., 2011), and of their dynamics during atmospheric transport (Beddows et al., 2014), but urban studies have been limited. There has been much research on emissions from road transport (Zhu et al., 2002a, b; Kumar et al., 2011), with some attention given to shipping (Gonzalez et al., 2011) and to general modelling of sources (Posser and Pandis, 2015). However, most urban measurement studies have been limited to a single site (Morawska et al., 1998; Wang et al., 2011; Brines et al., 2015), although in a few instances more sites have been considered (Karl et al., 2016) but not as part of a concerted campaign.
Within this study, particle number size distributions were measured simultaneously by electrical mobility spectrometers at five separate sites across London and the size distributions are compared with a view to gaining a better understanding of the sources and processes affecting particles in the urban atmosphere.
Data were collected from 27 January 2017 to 16 February 2017 as part of the second campaign of the FASTER project. Data recovery was high (100 %, or close) at all sites except Westminster University, where good SMPS (scanning mobility particle sizer) data were collected on only 3 days, 30 and 31 January and 1 February 2017.
Study area locations
Data were collected at five sampling sites in total, three of which were
established specifically for the FASTER campaign, Westminster University,
Regent's University and BT Tower. The other two sites (London Marylebone
Road and London North Kensington) collect data as part of the national
Automatic Urban and Rural Network. The site locations (seen in Fig. 1) and
characteristics are as follows:
The instruments (Table 1) were operated according to Wiedensohler et
al. (2012) guidelines, with the omission of a dryer at three sites (discussed
later), and calibrated and intercompared both before and after the sampling
campaign. Small correction factors (
Location sites of instruments during the campaign. Mean sea level (m.s.l.), above ground level (a.g.l.), condensation particle counter (CPC), scanning mobility particle sizer (SMPS).
Note: The SMPS size ranges are given in Sect. 2.2. The lower size cuts
(
It was not possible to use identical SMPS systems at each site. The variants used are shown in Table 1. We expect little difference between the long column classifiers (TSI 3081) used at all sites but with different platforms (TSI 3080 and TSI 3082) and CPCs (TSI 3775 and 3776). Differences are expected to be minimal as platform-specific software was used to invert the data and both the CPCs are butanol-based, with only slightly different lower cut-points which were well outside of the range of measured particles. At the Regent's University site, both a long DMA (3081) and short column DMA (3085) were utilised and the data were merged to give a single continuous size distribution from 6 to 650 nm. A possible cause of divergence is the fact that two of the sites (Marylebone Road and North Kensington) used diffusion dryers according to the EUSAAR/ACTRIS Protocol. The dryers were tested when installed and showed very low particle losses (less than 5 %) and no significant change to particle size distributions (NPL, 2010). The dryer may, however, affect the particle size distribution due to the hygroscopicity of certain kinds of particles. Vu et al. (2015b) reviewed hygroscopic growth factors for submicron aerosols from different sources. Their data are difficult to extrapolate to this study as measurements of hygroscopic growth are typically made at very high relative humidities, normally around 90 %. Even at 99.5 % relative humidity, the growth of particles of less than 100 nm sampled from the atmosphere is relatively low (Vu et al., 2015b). Consequently, a reduction in humidity from 88 % typical of the campaign to the values of 30–40 % achieved in the dryer would be expected to have only a small effect on particle sizes, especially as fresh traffic-generated particles which comprise a large proportion of the sub-micrometre particulate matter in the urban atmosphere are hydrophobic and therefore undergo zero or very limited growth in humid atmospheres.
Wind speed and direction data were taken from Heathrow Airport to the west
of London to reflect the synoptic flow minimally affected by local building
effects. At the start of the campaign (27 January 2017) the wind direction
was easterly and moved to southerly by 29 January, briefly passing through
northerly before returning to a southerly circulation between 31 January and
3 February. During this time, wind speeds were typically around 4 m s
The mixed layer heights (MLHs) were determined from Vaisala CL31 ceilometer data collected at the Marylebone Road site (Fig. 1, Table 1). The observed 15 s (10 m gates) aerosol attenuated backscatter profiles were pre-processed (Kotthaus et al., 2016) prior to using the CABAM algorithm (Kotthaus and Grimmond, 2018) to determine 15 min intervals of MLH. The multiple aerosol layers (e.g. nocturnal residual layers) in the atmosphere are detected (Kotthaus and Grimmond, 2018; Kotthaus et al., 2018). Here the lowest detected layer is analysed. At times the MLH cannot be detected (e.g. during rain or very weak gradients in attenuated backscatter), but a residual layer might still be indicated. The ceilometer detects periods of precipitation, including events that may not be recorded by ground-based stations (e.g. insufficient to trigger a tipping bucket rain gauge).
During the campaign the observed MLH varied from a daily minimum of 45 m a.g.l. to a daily maximum of 1312 m a.g.l. with an overall 15 min average (median) of 421 (382) m a.g.l. The daily average (median) maximum MLH was 777 (695) and minimum was 194 (197) m a.g.l. The daily range and the amount of data available per day are shown in Fig. S1 in the Supplement.
Modes were fitted to the 15 min data obtained at Marylebone Road and Regent's
and Westminster Universities using curve fitting and data analysis software
“Fityk (version 1.3.1)” developed by Wojdyr (2010). In the present
analysis, a standard peak function (Eq. 1) was used to disaggregate the
size distributions into lognormal modes:
By fitting a combination of
A time series of total particle number concentrations from the SMPS instruments appears in Fig. 2. A strong diurnal variation is seen at all sites and is exemplified by the average daily variation shown in Fig. 3.
Time series of total particle number count from the SMPS instruments at the five sites (Fig. 1, Table 1) over the campaign period.
The data stratified by the wind direction measured at London Heathrow
airport (LHR) (Fig. 4) were used to perform the modal analysis. The log-normal modes
fit to the size distribution were used to provide insights into
the separate modes contributing to a measured size distribution. Although
most measurements could be fit with three separate modes some distributions
were best fit with only two modes. An example of a three-mode fit of a size
distribution from North Kensington appears in the data for the
270
Campaign-average diurnal variation of particle number counts derived from the SMPS instruments with median (line) and inter-quartile range (shading) shown.
Average particle number size distributions stratified by
45
The Marylebone Road sampling site is located in a street canyon with heavy traffic
(approx. 80 000 vehicles per day). The canyon is aligned
almost east–west and the sampling site is at kerbside on the southern side
of the street. The canyon has a height-to-width ratio of
Lognormal modes fitted to the average particle size spectrum at
North Kensington for wind direction sector 270
A schematic diagram of the wind flows in the street canyon of Marylebone Road (six traffic lanes) during southerly and northerly winds. The orange marker represents the MR sampling site and red marker represents the WM sampling site.
The Westminster University sampling site is 22 m higher and slightly
displaced (
The North Kensington site is widely taken as representative of the background air pollution climate in central London (Bigi and Harrison, 2010; Bohnenstengel et al., 2015). At this site, the size of the first mode in the size distributions is remarkably constant at 22–26 nm, which is slightly larger than that observed at Marylebone Road. The second mode is also less variable than at most other sites and broadly within the range of the second mode sizes at Marylebone Road (see Table S1). The third mode is highly variable in size with wind direction but again broadly comparable to the data from Marylebone Road. The Beddows et al. (2015) positive matrix factorization of particle number size distributions data from this site identified four factors contributing to the particle number size distributions: a secondary component accounting for 4.4 % of particle number with a mode at around 250 nm, an urban background factor (43 % of particle number) peaking at around 50 nm, a traffic component (44.8 % of particle number) peaking at around 30 nm and a regional nucleation component (7.8 % of particle number) peaking at 20 nm. The regional nucleation component showed a strong seasonality with greatest prevalence in the summer months and is thought unlikely to have contributed significantly during the period of this campaign. This was a winter campaign without clear evidence of nucleation leading to new particle formation at any of the sites. A subsequent paper has investigated the factors influencing nucleation at three related sites, including North Kensington and Marylebone Road (Bousiotis et al., 2018). Consequently, the first mode observed in our current study is very comparable to the traffic mode observed by Beddows et al. (2015), and the second mode corresponds strongly to the urban background factor identified by Beddows et al. (2015) who associated this factor with aged traffic emissions and wood smoke, the latter of which is unlikely to have influenced the size distribution at Marylebone Road significantly.
Previous London work has shown the tendency of nucleation mode
traffic-generated particles sampled within Regent's Park to have shrunk by
evaporation at rates of on average 0.13 nm s
Under southerly flows the Regent's University site is downwind of Marylebone
Road (Fig. 1). The modal diameters measured at Regent's University in the
nucleation mode (Table S1) are clearly indicative of a shrinkage of particle
diameter for the wind sectors 180, 225
and 270
In our earlier studies of the evolution of particle sizes between Marylebone
Road and Regent's Park (Harrison et al., 2016), the nucleation mode in the
Marylebone Road size distributions lay between 20 and 24 nm (i.e. very similar
to this study). In Regent's Park this had reduced to within the range of
6–11 nm, with the largest sizes measured in the 0
Previous BT Tower site observations have reported loss of
Earlier studies have shown that particle number concentrations (
Average diurnal variations of total particle number count derived from the
condensation particle counters produced using the OpenAir software package
(Carslaw and Ropkins, 2012) appear in Fig. S2. At both Marylebone Road and
Westminster University, these show a peak occurs between midnight and
06:00 LT (local time) before reducing and then rising to a second peak in the afternoon. CPC
concentrations at these sites far exceed those at Regent's University and
the BT Tower, whereas integrated counts from the SMPS instruments were
considerably smaller and showed a diurnal variation broadly similar to that
expected for road traffic emissions (Fig. 3). While it is quite normal for
the CPC to give a higher count than the SMPS since it measures over a wider
size range and may have lower internal losses (although the SMPS data
analysis software corrects for internal losses), the ratio of CPC to SMPS is
in our experience (e.g. Shi et al., 2001) typically around 2, but this
value was significantly exceeded episodically, especially at Westminster
University (Fig. S3). The overall pattern of CPC to SMPS ratios (Fig. 7)
shows that some of the highest ratios were at Regent's University, with two
individual occasions exceeding 13. Some high peak values were observed at
Westminster University during the short SMPS time series. Wood burning is
recognised as an influential source of particles in London (Harrison et al,
2012b; Crilley et al., 2015) and has a diurnal profile with higher
concentrations typically at night. During the ClearfLo winter campaign the
BT Tower was influenced substantially by wood smoke irrespective of boundary
layer depth (Crilley et al., 2015). Since the BT Tower site was predominantly
within the mixed layer during the 2017 campaign (Fig. S1) and the
Time series (15 min) of ratio of total particle number counts,
To evaluate this phenomenon more closely, the black carbon data were
examined. These are typically taken as a good tracer of diesel exhaust, which
is expected to be the main source of the particle number count. The diurnal
variation in black carbon (Fig. S4) conformed reasonably well to that
expected for a traffic-generated pollutant, with Marylebone Road
concentrations far exceeding those at the other sites and showing a typical
traffic-associated pattern. The ratio of particle number (derived from the CPC) to black carbon (Fig. S5) shows huge diurnal variability similar to that seen
in the ratio of particle number count from the CPC to that derived from the
SMPS. We infer from this behaviour that a large number of particles smaller
than the lower limit of the SMPS and above the lower limit of the CPC
(i.e. 2.5–14.9 nm for the 3776 instrument at Westminster University and Regent's
University; 4–14.9 nm for 3775 instrument at BT Tower; and 3–16.55 nm for
3025 instrument at Marylebone Road) were present in the atmosphere. Both the
mean ratio of CPC to SMPS (Fig. S3) and CPC to black carbon (Fig. S6)
have ratios that are greatest in the early morning (midnight to 06:00 LT). This
is unexpected for the
Such behaviour is somewhat unexpected and a review of papers in which vertical gradients in particle number count have been measured above roadside sites showed no earlier evidence of such behaviour (Lingard et al., 2006; Agus et al., 2007; Nikolova et al., 2011; Ketzel et al., 2003; Longley et al., 2003; Kumar et al., 2008a, b, 2009; Li et al., 2007; Vakeva et al., 1999; Zhu et al., 2002b; Wehner et al., 2002). However, evidence is seen in some of Villa et al.'s (2017) observations that particle number count increased with height up to around 10 m above a multi-lane highway. The authors reported this unexpected pattern for some ascents and descents and attributed it to exhaust tubes of heavy-duty trucks tending to project vertically upwards and to be located at a height of several metres above ground. They suggest this is not the case in urban canyons.
Another possibility arises from the report of Rönkkö et al. (2017)
that large numbers of sub-4 nm particles are observed in the exhaust of some
diesel engines and the observation by Nosko et al. (2017) of substantial
numbers of similarly sized particles amongst emissions from brake wear.
Kontkanan et al. (2017) reported observations of sub-3 nm particles from
many sites, the highest concentrations being in urban locations. The diurnal
and regional variations did not relate clearly to photochemistry and it was
concluded that sub-3 nm particle concentrations are affected by
anthropogenic sources of precursor vapours. The correlation of sub-3 nm
particle concentrations in Helsinki with nitrogen oxides suggested a link
with traffic emissions. Shi et al. (2001) measured particles of
Support for our observations also comes from the very detailed measurement and modelling study of Choi and Paulson (2016). Measuring particle number size distribution downwind of a major highway, they found a positive anomaly in particle number within the first 60 m of the plume peak, as the peak for the small particles appeared further downwind than the peak in accumulation mode particles. They attributed this to growth of unmeasured sub-5.6 nm particles into the smallest measurable size range and suggested condensational growth or self-coagulation as the mechanism (Choi and Paulson, 2016). Kerminen et al. (2007) measuring near a major road in Helsinki reported particle growth by condensation to be a dominant process during the road-to-ambient evolution stage at night-time in winter. They inferred that under such conditions (low wind speeds with a temperature inversion), traffic-generated particle numbers were enhanced and could affect submicron particle number concentrations over large areas around major roads. The distance scales for such processes in both studies (Choi and Paulson, 2016; Kerminen et al., 2007) were within 100 m of the source under the conditions of measurement but might conceivably extend over greater distance scales. Similar processes of particle evolution within an aircraft exhaust plume have been reported by Timko et al. (2013).
Pushpawela et al. (2018) report a phenomenon of hygroscopic particle growth at night-time, which can potentially be mistaken for new particle formation. This phenomenon was observed between 0.5 and 5.0 h after sunset, peaking at 3.5 h (Pushpawela et al., 2018). This would not appear to explain our observations, where the peak in N–SMPS and N–BC particle number plots (Figs. S2 and S5) is greatest at 03:00–04:00 LT, which in London in winter is some 10–11 h after sunset. Additionally, such a phenomenon would be expected to be unrelated to local traffic emissions, and hence more uniform across the various sites.
Figure 2 shows the time series of particle concentrations from the SMPS instruments throughout the campaign. Clearly, as expected, the Marylebone Road site shows the highest concentrations through the campaign period due to its proximity to the road traffic source. The other sites tend to track one another quite closely with no consistent ranking of concentrations. There are periods such as 1 to 3 February when Regent's University well exceeds North Kensington, but at other times, they are very similar (e.g. 10–12 February), or periods when North Kensington exceeds Regent's University (e.g. 7 February) but these are few. In the former period (1–3 February), winds were southerly and concentrations at Regent's University would be enhanced by passage of air across central London, including Marylebone Road. In the situation where concentrations were similar (10–12 February), winds were in the northerly sector, giving relatively low concentrations at all sites, and rather little spatial variation. The temporal pattern at all sites showed substantial similarity overall (Fig. 2), including diurnal patterns (Fig. 3), although the magnitude of concentrations varied.
Time series (15 min) of total particle number count from the CPC instruments located at four sites over the campaign period.
A time series of CPC particle number concentrations (Fig. 8) showed that under most conditions, the number count was lowest at the BT Tower site, and that the number count at Westminster University frequently exceeded that at Marylebone Road, with Regent's University lower, but above the concentration at the BT Tower (Fig. 8). During the period of northerly winds (8–12 February), all sites showed low concentrations in the SMPS data, with Regent's University and BT Tower similar for much of the time (Fig. 2). The highest CPC count concentrations during the latter were measured at Westminster University (Fig. 9), which was downwind of Marylebone Road at those times. The similarity seen between Westminster University and Marylebone Road for much of the campaign, with concentrations far in excess of those at BT Tower, is strongly suggestive of continuing particle growth into the size range 2.5–14.9 nm at Westminster University, with re-evaporation occurring before reaching the elevated BT Tower site, as previously observed by Dall'Osto et al. (2011). Elevations in N–BC data were seen at the BT Tower site (Figs. S4 and S6) but these occurred mainly during the morning rush hour period, presumably due to fresh traffic emissions, rather than overnight as at the other sites (Fig. S6).
Time series (15 min) of
Figure 2 suggests that vertical gradients between the proximate Regent's
University and BT Tower sites were small in SMPS count (Fig. 2), but at
certain times were substantial in the CPC count (Fig. 9). The particle
size distributions measured at the BT Tower (Fig. 4d) differ from
Marylebone Road and North Kensington (Fig. 4a and b) in having no obvious
mode in the nucleation size range at 20–30 nm, a feature shared with
Regent's University (Fig. 4c). Only during westerly winds (270
The size distributions have also been analysed according to mixed layer
height, determined by ceilometer (Kotthaus and Grimmond, 2018). Both
Marylebone Road (Fig. S7) and Regent's University (Fig. S8) have the
highest concentrations associated with the deepest MLH class (
Unfortunately, a full dataset for the Westminster University site was only
collected over the period 30 January to 1 February due to a
late set-up of the instrument and a malfunction after 1 February.
This period, however, merits closer examination as it is the only period where
SMPS data were available for all three sites. For much of the time the SMPS
data for the Westminster University site looks surprisingly similar to that
of the Marylebone Road site despite the former being on the rooftop and the
latter being within the street canyon. A detailed analysis hour by hour
showed that out of 51 hourly observations, in 23 the amplitude of the mode
(dN
In order to gain further insight, the time series of observations were
plotted for this period and appear in Fig. 9. The SMPS-integrated number
counts shown in Fig. 9a show a remarkable similarity between Marylebone
Road, Westminster University and Regent's University. For the first 2
days, Regent's University concentrations are lower than those from the other
two sites, although on the third day they are very similar to those at
Westminster University. On the first and last days, the peak concentrations
at Marylebone Road exceed those at Westminster University but on the middle
day (31 January) the differences between these two sites are very
small. The CPC particle number counts shown in Fig. 9b are very similar
to those at Marylebone Road on the first and last day but exceed those at
Marylebone Road on 31 January. Concentrations at Regent's University
are typically only around half or less of those measured at Westminster
University. The magnitude of the CPC concentrations peaking at over 40 000 cm
However, the black carbon data (Fig. 9c) have daytime concentrations at Marylebone Road that far exceed those at Westminster University and Regent's University, the latter sites tracking each other and having very similar concentrations. Since black carbon can be viewed as a conserved tracer of vehicle emissions over these small time and distance scales, the inference is that particle production must be continuing as the vehicle exhaust mixes upwards from the street canyon Marylebone Road site to the Westminster University rooftop site. The southerly wind directions likely associated with upward flow on the Westminster University canyon wall (Fig. 6) would carry vehicle exhaust past the Marylebone Road measurement station (south side of the road).
Air leaving the canyon and being entrained by the complex building roof flows could expose the Westminster University sampler to air exiting the street canyon and to the general flow towards Regent's University site (Figs. 6 and 1). Such behaviour is consistent with the observations of particle growth in the sub-SMPS size ranges reported in the previous section, extending into the SMPS size range. This is similar to behaviour observed by Kerminen et al. (2007) in Helsinki, who observed not only possible evaporation of some particles in the 7–30 nm range, but also an apparent growth of nucleation mode particles into the 30–63 nm size range between sampling points at 9 and 65 m downwind of a highway. The results in Fig. 9 are suggestive of a substantial growth of nuclei into the range of the CPC at Westminster University.
The measurement of particle number size distributions in the atmosphere is resource intensive and there have been rather few studies in which more than two samplers have been operated within a city. Typically if there are two sites, one is a traffic-influenced site and the other urban background. In this study, data have been collected at a total of five sites, although unfortunately the dataset from the Westminster University site is limited to only a few days. Nonetheless, the dataset allows some deep insights into the spatial distribution of particle sizes and number counts not only horizontally but also in the vertical dimension. Not unexpectedly, concentrations of particles at the street canyon Marylebone Road site considerably exceed concentrations at other sites, but there are nonetheless considerable similarities in diurnal profiles and the magnitude of concentrations at the other, background sites.
One of the main motivating factors for this study was to confirm earlier
observations of shrinkage of the nucleation mode particles between traffic
emissions on Marylebone Road and the downwind site at Regent's University
within Regent's Park. Particle shrinkage was observed within the current
study although at a slower mean rate (0.04 nm s
Although the phenomenon of particle shrinkage had been seen in earlier work,
there were two further major observations made in the current study which
were not anticipated. The first was the clear influence of a major source
to the west of London, almost certainly Heathrow Airport, upon
concentrations of nucleation mode particles. The association of an enhanced
nucleation mode in the 270 or 225
The other observation which was wholly unexpected was of the very poor
relationship between total particle numbers measured by the scanning
mobility particle sizers and the total particle numbers measured by
co-located condensation particle counters. While both the SMPS counts and
co-located black carbon measurements show a typical road traffic diurnal
profile, the CPC data show a quite different diurnal profile peaking at
night. This is most evident in the ratios of
These very abundant particles within the 2.5–15 nm range are likely to prove ephemeral as they would be expected to re-evaporate as the air mass dilutes away from the source. However, the health effects of exposure to particles within this range are poorly known and no recommendation can be given as to whether health-related studies would be best to measure the particle size range covered by the SMPS as is most typically performed at present, or whether CPC data going down to smaller particles sizes would be more appropriate.
There are some additional general conclusions from the work. Firstly the results demonstrate the dynamic behaviour of traffic-generated (and other) particles within the urban atmosphere. Our earlier paper (Dall'Osto et al., 2011) referred to “remarkable dynamics”, and further remarkable dynamic processes have been observed in the current study. Secondly, as this work has revealed sources and processes that were not originally anticipated, although with the benefit of hindsight it might have been possible to predict them, there is clearly a need for further detailed observational studies of the behaviour of sub-100 nm particles within the urban atmosphere.
Data supporting this publication are openly available from
the UBIRA eData repository at
The supplement related to this article is available online at:
DB, MA, JB and RX carried out the field measurements of particle size distributions, SK and SG collected and interpreted the ceilometer data, and DB and AS carried out data analyses. RH led the project and drafted the paper, with all co-authors contributing to subsequent enhancements.
The authors declare that they have no conflict of interest.
The authors are grateful to the management and staff of Westminster University, Regent's University and British Telecom for access to their buildings for air sampling. They also express gratitude to the National Centre for Atmospheric Science (NCAS) for the loan of sampling instruments, and to Paul Williams (NCAS) for facilitating the instrument intercomparison. The operation of the ceilometers were supported by NERC ClearfLo, NERC AirPro, Newton Fund/Met Office CSSP (SG, SK) and University of Reading. We acknowledge the support of KCL LAQN for the instrument sites and support and the Reading Urban Micromet group for maintaining the instruments, notably Elliott Warren and Kjell zum Berge. The work was funded by the European Research Council (ERC-2012-AdG, proposal no. 320821) and the UK Natural Environment Research Council (R8/H12/83/011) and a NCAS studentship (to JB). Edited by: Veli-Matti Kerminen Reviewed by: two anonymous referees