Introduction
With frequent occurrence of haze episodes, the megacity of Beijing is faced
with severe air pollution problems, as indicated by high concentrations of
ambient aerosol particles. For example, the annual average concentration of
PM2.5 was 80.6 µg m-3 in 2015, which is more than twice the
China National Ambient Air Quality Standard (35 µg m-3 as an
annual average)
(http://www.bjepb.gov.cn/bjepb/413526/413663/413717/413719/index.html). Fine
particles can significantly reduce atmospheric visibility, exert harmful
effects on public health, and even have potential impacts on regional and
global climate. As a result, extensive efforts have been devoted to
characterize the sources, formation mechanisms, and evolution processes of
aerosol particles in recent years (Ma et al., 2012; Takegawa et al.,
2009; Sun et al., 2010, 2014,
2016b). Among these studies, particle number concentrations are one of the
greatest concerns because particles can rapidly grow from a few nanometers
to tens and even hundreds of nanometers in a short time, and hence play a
significant role in haze formation (Guo et al., 2014). However, our
understanding of the formation and growth of aerosol particles is not
complete, particularly in highly polluted environments (Kulmala et al.,
2016).
In the past decades, extensive studies have been conducted to characterize
particle number size distributions in Beijing at ground level (Wehner et
al., 2004; Yue et al., 2009; Wu et al., 2011; Gao et al., 2012; Wang et al.,
2013b). The continuous measurements of aerosol number size distributions
from 3 nm to 10 µm within the city area of Beijing in spring
indicated a high variability in number concentrations, and the variations
were substantially different among dust storm, clean, and polluted periods
(Wehner et al., 2004). Yue et al. (2009) also
found a clear shift of maximum diameter from 60 nm on clean days to 80 nm
during polluted days. Most previous studies focused on new particle
formation and growth events (NPEs) (Wehner et al., 2004; Wu et al.,
2011; Gao et al., 2012; Zhang et al., 2011). While new particle formation (NPF) events are mostly observed under
conditions with low relative humidity and clean air masses (Wu et al.,
2007; Wehner et al., 2004), particle growth events are strongly associated
with high relative humidity (Gao et al., 2012). The roles of
chemical species in NPEs in Beijing were also explored in several studies.
For example, organics were found to be the dominant species of PM1
during new particle formation events in summer in Beijing
(Zhang et al., 2011) and likely played a major role
in NPF and growth, although
sulfuric is also important as well (Yue et al., 2009).
However, most of these studies were conducted at a ground site that is
subject to the influences of multiple local sources, e.g., traffic and
cooking emissions. Indeed, the source apportionment of particle numbers with
positive matrix factorization showed significant contributions of traffic
emissions and combustion sources to the total number concentration (Wang
et al., 2013b; Liu et al., 2014). Therefore, measurements of size-resolved
number concentrations at high altitude with less local cooking and traffic
influences are essential for elucidating the NPF and growth mechanisms, and
also the role of regional transport in haze formation.
During this study period, strict emission controls were implemented in
Beijing and the surrounding regions, e.g., Hebei, Tianjin, and Shandong, from 20
August to 3 September to ensure good air quality during the China
Victory Day (V-day) parade on 3 September 2015. The control measures, such as
restricting the number of vehicles, shutting down factories and power
plants, stopping construction activities, etc., were even stricter than
those implemented during the Asia-Pacific Economic Cooperation (APEC) summit
in 2014 (Sun et al., 2016a). Several studies have addressed the impacts
of regional emission controls on aerosol composition and gaseous species
(Li et al., 2016; Han et al., 2016; Zhao et al., 2017). The results are
overall consistent, showing significant reductions in most aerosol and
gaseous species during the control period (CP, 22 August–3 September). A
recent study comparing the number size distributions with those during
the same period in 2010–2013 at a rural site in Beijing illustrated the most
reductions in accumulation-mode particles and condensation sink (CS) during
the V-day period (Shen et al., 2016). Despite this, our understanding of
the impacts of emission controls on particle number size distributions is
far from complete.
Here, we conducted the first simultaneous measurements of particle number
size distributions at two different heights, i.e., ground level and 260 m, within the city area of Beijing from 22 August to 30 September. This study
is unique because it provides an experimental opportunity to investigate the
vertical differences and processes of particle number size distributions and
also the impacts of regional emission controls. The size-resolved particle
number concentrations, diurnal variations, particle growth rates, and their
relationships with aerosol composition at ground level and 260 m are compared
in detail, and the impacts of emission controls on particle number
concentrations in different sizes are elucidated. In addition, the sources
of particle numbers at the two different heights are investigated with
positive matrix factorization.
The geometric mean diameter (GMD) of average particle number size
distribution for different periods at ground level and 260 m. Also shown are
GMDs of three modes from log-normal fitting.
GMD
Entire study
Three modes
260 m
Ground
260 m
Ground
Entire study
88
45
27
44
116
24
41
111
Control period
57
43
27
48
104
24
46
150
Non-control period
106
47
27
43
119
23
40
102
Clean
79
47
27
45
112
23
41
106
Polluted
131
47
52
113
188
36
96
244
Experimental method
Sampling and data analysis
The sampling site is located at the Tower Branch of the Institute of Atmospheric
Physics, Chinese Academy of Sciences, between the north third and fourth ring
roads in Beijing. Two scanning mobility particle sizers (SMPSs) were deployed
for simultaneous measurements of particle number size distributions at
ground level and 260 m on the Beijing 325 m meteorological tower. At 260 m,
the size-resolved particle number concentration (15–685 nm) was measured
in situ by a condensation particle counter (TSI, 3775) equipped with a long
differential mobility analyzer (TSI, 3081A). The time resolution is 5 min. Comparatively, an SMPS as part of an unattended multifunctional
hygroscopicity tandem differential mobility analyzer (H-TDMA) developed by
the Guangzhou Institute of Tropical and Marine Meteorology, China
Meteorological Administration (ITMM, CMA) was used to measure particle
number concentrations (10–400 nm) at ground level. A detailed description
of the H-TDMA was given in Tan et al. (2013). According to previous
comparisons of particle number size distributions between different SMPSs or
differential mobility particle sizers, the measurement uncertainties
between 20 and 200 nm can be ∼ 10 %, and even larger for
particles outside this range (Wiedensohler et al., 2012).
The non-refractory submicron aerosol (NR-PM1) species, including
organics (Org), sulfate (SO4), nitrate (NO3), ammonium (NH4),
and chloride (Chl), were measured at ground level by an Aerodyne
high-resolution time-of-flight aerosol mass spectrometer (HR-AMS) and at
260 m by an aerosol chemical speciation monitor (ACSM). Co-located
black carbon (BC) was measured by a seven-wavelength (AE33) and a
two-wavelength Aethalometer (AE22, Magee Scientific Corp.) at 260 m and
ground level, respectively. The meteorological variables, including wind
speed (WS), wind direction (WD), relative humidity (RH), and temperature
(T), were obtained from the measurements on the meteorological tower. The
operations of the HR-AMS, ACSM, and Aethalometers and subsequent data
analysis are detailed in Sun et al. (2015b) and Zhao et al. (2017).
All the data in this study are reported in Beijing local time (UTC+8 h).
Figure S1 in the Supplement shows a comparison of the total PM1 mass (NR-PM1+BC) with that derived from the SMPS measurements at ground level and 260 m.
The particle number concentrations between 15 and 400 nm were converted
to mass concentrations using chemically resolved particle density
(Salcedo et al., 2006). As shown in Fig. S1, the time
series of PM1 was highly correlated with that from SMPS measurements at
both ground level (r2=0.94) and 260 m (r2=0.95). We also
noticed some differences in the regression slopes, which are 0.44 and 0.66
at ground level and 260 m, respectively. The reasons are not very clear yet,
but these differences are likely due to the different size distributions at the two different
heights (Sect. 3.1).
Particle growth rates and condensation sink
The particle growth rates (GRs) at ground level and 260 m were calculated
using Eq. (1).
GR=ΔDmΔt,
where Dm is the geometric mean diameter from the log-normal fitting of
each size distribution and ΔDm is the increase in diameter during
the growth period of Δt.
Average particle number size distributions during the (a) control
period, (b) non-control period, and (c) the entire study at ground level
(orange lines) and 260 m (green lines). Panel (d) shows the time series of
meteorological parameters of relative humidity (RH) and temperature (T). Panels (e)
and (g) are the particle number size distributions and condensation
sinks
(CSs) at 260 m and ground level, respectively. Panel (f) and (h) are the time
series of mass concentrations of PM1 species at 260 m and ground level,
respectively.
Condensation sink (CS), indicating how rapidly vapor molecules can condense
onto preexisting aerosols, is calculated using Eq. (2)
(Nieminen et al., 2010).
CS=2πD∑iβMiDp,iNi,
where D is the diffusion coefficient of the condensing vapor, Dp and N
are
the particle diameter and the corresponding number concentration, and
βM is the transitional regime correction factor expressed as
Eq. (3).
βM=Kn+1/1+0.377Kn+43α-1Kn2+43α-1Kn,
where α is assumed to be unity, and Kn is the Knudsen number. It
should be noted that the CS calculated on the basis of dry particle number
size distributions might be underestimated since ambient RH was not
considered (Reutter et al., 2009).
Comparisons of particle number concentrations between ground level
and 260 m for different size ranges, i.e., (a) small Aitken mode
(15–40 nm), (b) large Aitken mode (40–100 nm), (c) accumulation mode
(100–400 nm), and (d) the total number of particles (15–400 nm). The right-hand panels show
the scatter plots of the comparisons.
Source apportionment of size-resolved particle number
concentrations
Positive matrix factorization (PMF, 2.exe, v 4.2) was performed on the
size-resolved number concentrations (Ulbrich et al., 2009; Paatero and
Tapper, 1994) to resolve potential sources. In this study, the measurement
uncertainties were estimated using an equation-based approach that was
detailed in Ogulei et al. (2007). The required measurement errors
(σij) were first calculated using Eq. (4):
σij=C1×(Xij+Xj¯),
where C1 is a constant value assumed to be 0.01, Xij is the measured
particle number concentration, and Xj¯ is the arithmetic mean value for
jth size bin. The measurement uncertainties (Unc) were then calculated
with Eq. (5):
Uncij=σij+C2×Xij,
where σij is the estimated measurement error and C2 is a
constant value assumed to be 0.1. After a careful evaluation of the
PMF results, five factors and four factors were chosen at ground level and 260 m,
respectively. A more detailed diagnosis of PMF results is presented in
Figs. S2 and S3.
Results and discussion
Characterization of particle number size distributions
The temporal variations in size-resolved number concentrations and aerosol
species at ground level and 260 m are shown in Fig. 1. The size-resolved
particle number concentrations showed overall similar evolutionary patterns
between ground level and 260 m, and high number concentrations of large
particles were generally associated with correspondingly higher
concentrations of aerosol species, e.g., the periods of case 1 and case 2 in
Fig. 1. However, periods with substantially different number size
distributions were also observed. For example, we observed significantly
higher particle number concentrations at ground level than 260 m in the evening
on 26 August and 1 September due to the influences of local cooking
emissions. On average, the particle numbers showed a broader size
distribution at 260 m than ground level, peaking at approximately 85 and
45 nm, respectively (Fig. 1c). The log-normal distribution fitting further
illustrated three size modes at both ground level and 260 m. While the
second mode with geometric mean diameter (GMD) peaking at 41 nm accounted
for the largest number fraction at ground level (52 %), the largest mode
(GMD = 116 nm) dominated the total number of particles at 260 m,
accounting for 62 %. Such differences were likely due to the stronger
influences of local sources (e.g., cooking) with higher emissions of smaller
particles, and more influences of regional transport with large
aged particles at 260 m.
Figure 2 shows the comparisons of the total number concentrations (15–400 nm, N15-400) and those for
three modes, including small Aitken mode (15–40 nm, N15-40), large Aitken mode (40–100 nm, N40-100), and
accumulation mode (100–400 nm, N100-400) between ground level and
260 m. The variation trends in the total number concentrations at the two
heights tracked relatively well (r2=0.40, slope = 0.71), while
the average number concentration from 15 to 400 nm at 260 m (7473 ± 4324 cm-3) was
26 % lower than that (10 134 ± 4680 cm-3) at
ground level. The total particle number concentrations at ground level were
generally lower than those previously observed in Beijing mainly due to the
smaller size range measured in this study (Wang et al., 2013b; Yue et al.,
2009). The N15-400 ratio of 260 m to
ground (R260m/ground) varied dramatically throughout the entire
study,
with the daily average ranging from 0.42 to 1.10. In contrast, the total
volume concentrations showed much better correlations between ground level
and 260 m (r2=0.89) and the average ratio was close to 1. Such
differences were mainly caused by the different contributions of different
mode particles to the number and volume concentrations.
Summary of average number concentrations of five factors for the
entire study, control period (CP), non-control period (NCP), and also the
change percentages ((CP-NCP)/NCP × 100).
F1
F2
F3
F4
F5
260 m
Ground
260 m
Ground
260 m
Ground
260 m
Ground
260 m
Ground
Entire study (cm-3)
867
695
–
2567
2066
3376
2859
2662
1412
801
Control period (cm-3)
1067
816
–
2586
2271
3314
2049
1619
489
357
Non-control period (cm-3)
771
621
–
2526
1967
3413
3249
3162
1856
1023
(CP-NCP)/NCP (%)
38 %
31 %
–
2 %
15 %
-3 %
-37 %
-49 %
-74 %
-65 %
Comparisons of particle volume concentrations between ground level
and 260 m for different size ranges, i.e., (a) small Aitken mode
(15–40 nm), (b) large Aitken mode (40–100 nm), (c) accumulation mode
(100–400 nm), and (d) the total number of particles (15–400 nm). Right figure show
the scatter plots of the comparisons.
The correlations of particle number and volume concentrations between ground
level and 260 m varied substantially for different mode particles. As shown
in Fig. 2a, the small Aitken-mode particles were correlated between the two
heights (r2=0.66), indicating their common sources that are related
to new particle formation. However, the average number concentration at 260 m (1382 ± 1281 cm-3) was only approximately 40 % of that at the
ground level (3379 ± 2232 cm-3), and the daily average ratio of
260 m to ground level for N15-40 varied from 0.91 to 0.51. These results
illustrated additional sources for small Aitken particles at ground level.
Indeed, pronounced peaks for N15-40 were often observed in the evening,
likely indicating the influences of local emissions, e.g., cooking and
traffic emissions. The large Aitken-mode particles showed the worst
correlation between ground level and 260 m (r2=0.40, slope =
0.70), although the average number concentrations were comparable (4188 vs.
3233 cm-3). These results suggested that the sources of large Aitken-mode
particles were quite different between ground level and 260 m. For example,
the diurnal cycle of large Aitken-mode particles at ground level was
remarkably similar to that of organic cooking aerosols (Zhao
et al., 2017), likely indicating a large source contribution from cooking
emissions. Compared with Aitken particles, the number and volume
concentrations of accumulation-mode particles were well correlated between
the two heights (r2=0.85 and 0.91, respectively). While the average
number concentration at 260 m was 11 % higher than that at ground level,
the volume concentration was close. Moreover, the temporal variations in
accumulation-mode particles tracked well with those of secondary inorganic
species that were mainly formed over a regional scale. Our results indicate
that accumulation-mode particles were likely dominantly from regional
transport and relatively homogeneously distributed across different heights.
The different vertical ratios between number and volume concentrations
suggest that the particle size distributions were slightly different between
ground level and 260 m.
Average number and volume concentrations and CS at (a) 260 m and
(b) ground level for the entire study and four different periods. Panel (c) shows
the box plots of the ratios of 260 m to ground level. The volume
concentrations of small and large Aitken-mode particles are enhanced by factors of
100 and 10, respectively, for clarity.
The regional emission control and meteorological conditions can have
significant impacts on particle number size distributions. As shown in Fig. 1a and b,
the GMD of number size distributions peaked at 57 nm at 260 m and
43 nm at ground level during the control period, and the
average size distribution showed three similar modes between the two
heights. In contrast, the size distributions had substantial changes after
the control period that were characterized by much broader distributions
and a clear shift from smaller to larger particles at both ground level and
260 m. For example, the GMD of particle number distributions was 106 nm at
260 m, which was much larger than that during the control period, and
the largest mode consistently dominated the total number of particles, on
average accounting for 68 %. Figure 4 shows a comparison of average number
and volume concentration between control and non-control periods for three-mode particles. While the average total number concentrations during the control
period were lower than those during non-control periods (6139 vs. 8116 cm-3 at 260 m and 8708 vs. 10 824 cm-3 at ground level), the small
and large Aitken-mode particles were comparable between control and
non-control periods. As a result, the decreases in total number
concentrations were mainly caused by the changes in accumulation-mode
particles, which were decreased by 53 % at 260 m and 52 % at ground level
during the control period. Our results illustrate that regional emission
control has a large impact on accumulation-mode particles, while the
influences on Aitken-mode particles were small. One of the major reasons is
that emission controls substantially decrease the gas precursors (e.g., SO2 and
NOx) and PM2.5 mass concentrations, and hence
suppress the growth of particles. This is also consistent
with the large decreases in CS by 48 % at 260 m and
45 % at ground level during the control period (Fig. 4a and b). In
addition to regional emission controls, we also found that the dominant
northerly winds likely played an important role in decreasing the PM during
the control period (Zhao et al., 2017). In contrast, the number
concentrations of small particles were relatively comparable due to more
frequent new particle formation events during the control period. To better
evaluate the impacts of regional emission controls, cluster analysis with
hourly back trajectories were performed on the entire dataset with an
exclusion of precipitation days. As shown in Fig. S4, accumulation-mode
particles during the control period showed the largest reductions for
clusters 1 and 2 (39 and 42 %, respectively), while the large Aitken
particles had small changes and the small Aitken ones even showed a large
increase (43 %) for cluster 1. These results further support our
conclusion above.
Average diurnal variations in particle number size distributions
at (a) 260 m and (b) ground level and (c) the ratios of 260 m to ground
level for the entire study. Panel (d) shows the ratios of particle number
concentrations at 260 m to those at ground level as functions of particle
size.
The diurnal cycles of particle number concentrations at 260 m and
ground level and the ratios of 260 m to ground level for different size ranges,
i.e., (a) 15–400 nm (N15-400), (b) small Aitken mode (N15-40),
(c) large Aitken mode (N40-100), and (d) accumulation mode
(N100-400).
We also compared the particle number size distributions between polluted
(PM2.5 > 75 µg m-3) and clean days
(PM2.5 < 75 µg m-3) after the control period. As shown in Fig. S5, the average size distribution on polluted days at ground level showed a
clear three-mode distribution, peaking at 36, 96, and 244 nm.
The GMD of the three modes was ubiquitously larger than those (23, 41, and 106 nm) observed during clean days. While the average total number concentration
increased from 10 258 (±4676) cm-3 during clean periods to
12 156 (±4406) cm-3 on polluted days (Fig. 4), we observed
comparable concentrations for small and large Aitken-mode particles.
Therefore, the increase in total number concentration was mainly caused by
the accumulation-mode particles, which increased by 90 % during the
polluted days. These results illustrate the different roles of different
mode particles between clean and polluted days. Similarly, the average
particle number distribution showed a clear shift from a smaller size during
clean periods to a larger size on polluted days at 260 m, and the total number
concentration was increased by 53 % from 7006 (±4416) to 10 748
(±3615) cm-3. Again, the increase in total number concentration
was mainly due to the increase in accumulation-mode particles by 135 %.
Compared with the number concentrations, the increases in volume
concentrations for accumulation-mode particles were more significant on
polluted days, which on average were 174 and 212 % at ground level and
260 m, respectively. Indeed, the accumulation-mode particles accounted for
97 % of total volume concentrations at both ground level and 260 m,
elucidating their major roles in PM pollution. The average number ratios
between 260 m and ground level increased as a function of particle sizes
during both clean and polluted days. For example, the ratios increased from
0.4 to 0.9 for small and large Aitken-mode particles, and to 1.2 for
accumulation-mode particles on polluted days. These results are consistent
with our previous conclusion that smaller particles showed stronger vertical
gradients than larger particles. We also observed ubiquitously higher
R260m/ground on polluted days compared to clean periods, indicating larger
vertical gradients in both number and volume concentration during polluted
periods.
Diurnal variations
The average diurnal variations in particle number size distribution at
ground level and 260 m for the entire study are shown in Fig. 5. It is clear
that particle number size distributions show pronounced diurnal cycles that
were characterized by the lowest values in the early morning and subsequent
particle growth until midnight. During the growth period, the GMD increased
from 29 to 57 nm in 14 h at ground level, while it increased from 41 to
88 nm in 12 h at 260 m. After that, the GMD remained at relatively constant
levels of 70 and 100 nm at ground level and 260 m,
respectively. The ubiquitously lower GMD and lower growth rates at ground
level were likely due to the influences of local emissions that contained a
large number of small particles. Note that the changes in GMD were
significant at ground level after the control period, especially on polluted
days (Fig. S6), indicating that the diurnal evolution of particle number
size distributions at ground level is subject to multiple influences. In
contrast, the changes at 260 m were much smaller with a relatively
consistent mode peaking at ∼ 100 nm, indicating a more
constant particle source at higher altitudes.
The particle number ratios between 260 m and ground level depend strongly on
particle size. As shown in Fig. 5, R260m/ground increases rapidly
between 15 and 100 nm as particle size increases but is typically less
than 1. This is consistent with our previous conclusion that small
particles are more abundant at ground level due to the influences of local
emissions. R260m/ground increases continuously and reaches a maximum at
Dp=∼ 150 nm. One explanation is the faster
condensational and coagulable growth of small particles at 260 m than
at ground level. Another explanation is the enhanced regional transport of 100–200 nm particles at high altitude. This is consistent with the fact that
much higher R260m/ground was observed during polluted periods than clean
periods. R260m/ground decreased to less than 1 at Dp>250 nm, likely due to the deposition of large particles. Our results show
that the vertical differences in particle number concentrations varied
significantly as a function of size, which has important implications that
the health and climate effects of aerosol particles at different heights
could be substantially different.
The diurnal cycles of particle number and volume concentrations at 260 m and
ground level, as well as R260m/ground during different periods, are
illustrated in Figs. 6 and S7. Pronounced diurnal cycles with two clear
peaks at noon and in the evening were observed at both ground level and 260 m.
Further analysis highlights that these two peaks were driven by small and
large Aitken-mode particles (Fig. 6b and c), likely
representing two dominant sources of new particle formation and cooking
emissions, respectively. In comparison, the diurnal cycles of accumulation-mode particles were relatively flat, indicating that the sources were mostly
regional. Figure 6 shows that the total particle number concentration during
the control period was consistently lower than that after the control
period, particularly during the time period of 00:00–08:00. Such decreases
were mainly caused by accumulation-mode particles that were reduced by 32–67 % at ground level and 23–69 % at 260 m throughout
the day. In contrast, the diurnal cycle of small Aitken-mode particles was
substantially different, characterized by a prominent peak between
10:00–14:00 associated with new particle events and a second smaller
peak at nighttime due to the influences of local emissions. The particle
number concentration of the NPE peak during the control period was even
higher than that after the control period, while the difference at nighttime
was much smaller. These results suggest that regional emission controls
could increase the number of small particles while significantly decreasing accumulation-mode particles. One explanation is that the growth of small
particles was suppressed due to the lower concentrations of precursors and
PM loadings. The diurnal cycles of R260m/ground for different sizes were
overall similar during and after the control period. As shown in Fig. 6, the diurnal cycles of R260m/ground are all
characterized by clear daytime increases due to enhanced vertical mixing,
and subsequent decreases at nighttime due to more influences of local source
emissions at ground level.
Average diurnal evolution of particle number size distributions
and aerosol composition at (a, b) 260 m and (c, d) ground level for the new
particle growth events. The dashed lines in panels (a) and (c) are the diurnal cycles
of CS.
We also compared the diurnal cycles of particle number concentrations
between clean and polluted periods. Again, very different diurnal profiles
were observed for particles in different size ranges. While small Aitken-mode particles at 260 m showed clear daytime increases during both clean and
polluted periods, those at ground level varied more dramatically due
to the influences of multiple sources. Similarly, the total number of small
Aitken-mode particles was slightly lower during polluted periods compared to
clean periods. In contrast, the diurnal cycles of large Aitken-mode
particles were quite different between ground level and 260 m. While a
pronounced nighttime peak due to cooking influences was observed at ground
level, more diurnal peaks that were associated with different sources and
processes were observed at 260 m. The largest difference between clean and
polluted periods was observed during 00:00–08:00 at 260 m, while the difference was
much smaller at ground level. Such differences clearly indicate very
different vertical gradients between clean and polluted periods for large
Aitken-mode particles. Compared to Aitken-mode particles, the number
concentration of accumulation-mode particles during polluted periods was
more than a factor of ∼ 2–3 higher than those during clean periods.
These results suggest that the major difference of particle number
characteristics between clean and polluted periods is accumulation-mode
particles. In fact, the CS during polluted periods was nearly twice that of
the CS during clean periods (Fig. 4), which facilitated the growth of particles.
Chemistry of particle growth
New particle growth events were frequently observed during the entire
study at both ground level and 260 m. As shown in Fig. 7, the growth process
of particles at ground level started from approximately 09:00 until midnight,
with the GMD increasing from ∼ 22 to ∼ 60 nm.
These results were consistent with those previously observed at urban and rural
sites in Beijing (Wang et al., 2013a). Similarly,
the growth of particles started from ∼ 28 nm at 09:00 to
∼ 63 nm at midnight at 260 m. The growth of particles was
closely related to the diurnal cycle of CS, which showed a continuous
increase from early morning to midnight. Also, aerosol composition had
significant changes during the growth periods. As indicated in Fig. 7b and
d, the contribution of organics first showed an increase during the early
growth period between 08:00 and 12:00, while those of other chemical species
remained small. After 12:00, both organics and sulfate showed
increased contributions until 17:00. Although the increases in organics and
sulfate were partly due to the decreases in nitrate and chloride because of
the evaporative loss in the afternoon, our results likely indicate that
organics played an important role in the early stage of particle growth,
while both organics and sulfate are important in the subsequent growth.
(a) Time series of particle growth rates and corresponding average
chemical composition for selected particle growth events. Panels (b) and (c) show
the correlation of particle growth rates with the changes in the
concentration of organics (ΔOrg) at 260 m and ground level,
respectively. Panels (d) and (e) show the correlation of particle growth rates with
condensation sink at 260 m and ground level, respectively. The data points
in (b–e) are color coded by the mass concentration of sulfate (SO4),
and those with sulfate concentrations higher than 3 µg m-3
(ground level) and 2.5 µg m-3 (260 m) are marked as triangle
points.
Panels (a) and (b) show factor profiles of particle number size distributions
at 260 m and ground level, respectively. (c) Comparisons of the time series
of PMF factors at 260 m (dashed gray lines) and ground level (color-coded
lines).
We further calculated the particle GRs for each growth event
that lasted more than 3 h (Fig. 8). The particle GR varied from 1.4
to 7.5 nm h-1 at 260 m and from 1.5 to 6.1 nm h-1 at ground level, which generally falls within the range that was
reported previously in various environments, e.g., Beijing (Wu et al., 2007; Zhang et al.,
2011), Shangdianzi (Shen et al., 2011), Egbert
(Pierce et al., 2014), Marseille (Kulmala et al.,
2005), and New Delhi (Sarangi et al., 2015). Particle growth rates
strongly depend on temperature and the availability of condensable vapors.
Indeed, the particle GR in the study generally correlated well with CS
at both ground level and 260 m during periods with low sulfate
concentrations (Fig. 8d and e). The average particle GR was 3.6 nm h-1 at 260 m, which is slightly higher than 3.3 nm h-1 at ground
level, which is likely due to the lower temperature at high altitude. It is
interesting to note that GR was correlated with the change in organic
concentration (ΔOrg) at 260 m, and also correlated well with ΔOrg
during periods with low sulfate concentrations (e.g., < 3 µg m-3) at ground level, likely indicating a dominant role of organics in
particle
growth. As shown in Fig. 8a, high sulfate concentrations were generally
observed during polluted periods with high PM loadings, and correspondingly,
relatively higher GR was related to higher sulfate concentration. Our
results here suggest that the particle growth mechanisms could be different
between clean periods with a dominance of organics and polluted periods with
significantly enhanced sulfate.
Source apportionment
PMF analysis of size-resolved particle number concentrations was able to
identify four factors and five factors at 260 m and ground level, respectively (Fig. 9). The five-factor solution at 260 m yielded a split factor that cannot be
physically interpreted. The average number size distributions of factor 1
showed GMDs peaking at 20 and 27 nm at ground level and 260 m, respectively,
and the temporal variations were characterized by frequent sharp peaks on
most days (Fig. 9c). It is clear that this factor was associated with new
particle events. This is further supported by the pronounced diurnal cycles
showing rapid increases between 08:00 and 12:00, and a dominant source region
to the west (Fig. S9a), where clean air masses were prevalent. However, we
also noticed the differences in diurnal cycles between ground level and 260 m. For example, the diurnal cycle of factor 1 at the ground site showed two
peaks during morning and evening traffic hours, likely indicating the
influence of traffic emissions. In fact, the time series correlation between
the two heights was weak (r2=0.17), confirming that the sources of
factor 1 are not the same. The average particle number concentrations of
factor 1 were 816 and 1067 cm-3 at ground level and 260 m during the
control period. These values were 31 and 38 % higher than those after the
control period. One explanation is due to the increase in CS after the
control period, which facilitated the condensation and coagulation of small
particles. This result indicates that regional emission controls could
increase the number of nucleation mode particles by reducing PM loadings and
decreasing CS. Note that a higher number concentration of factor 1 during the
control period was also likely due to the more frequent new particle
formation events associated with prevailing northerly winds (Zhao et
al., 2017).
Factor 2 presented a size distribution peaking at ∼ 32 nm and
a distinct diurnal cycle with two comparable and pronounced peaks at noon
and in the evening. The diurnal cycle of factor 2 resembled that of organic aerosol from cooking that was widely reported in Beijing (Huang et al.,
2010; Sun et al., 2013; Zhang et al., 2016; Elser et al., 2016; Xu et al.,
2015). On average, this factor accounted for 25 % of the total particle
number concentration and had only a small difference (2 %) between
the control and non-control periods. This factor was likely dominantly
contributed by cooking emissions, although particle growth can partly explain
the high concentrations during the late afternoon. Factor 3 at ground level
showed a similar diurnal cycle to factor 2, yet the evening peak was much
higher than the noon peak. Such a diurnal profile was remarkably similar to that
of organic aerosol from cooking that was resolved from PMF analysis of organic aerosol during the same study
period (Zhao et al., 2017). Also, the particle number size
distribution of factor 3 was similar to that from cooking activities
(Buonanno et al., 2011). These results supported the conclusion that factor 3
was mainly from cooking emissions. Similar to factor 2, there was only a
small change (3 %) during and after the control period, consistent with
the fact that no control measures were implemented near our sampling site
during the control period. Compared to the ground site, factor 3 at 260 m also showed two pronounced peaks in the diurnal profile. However, the
nighttime peak was much smaller than that at ground level. This can be
explained by the significantly enhanced cooking emissions at nighttime at
ground level. However, the vertical mixing to high altitude was limited due to the fact that the
average number concentration at ground level was 3375 cm-3, which was
64 % higher than that at 260 m. This indicates stronger influences of local
cooking emissions on particle numbers at lower altitudes. This factor
moderately correlated between ground level and 260 m (r2=0.37),
indicating that cooking sources could also be different at different
altitudes, for example, more contributions from regional cooking emissions
at higher altitudes. In addition, factor 3 at 260 m was better correlated
with the sum of factor 2 and factor 3 at ground level (r2=0.40,
Fig. S8), further supporting that these three factors have similar sources.
More evidence is that factors 2 and 3 have the smallest influences from
regional emission control among all factors.
Factors 4 and 5 showed quite different temporal variations, but were
generally characterized by high concentrations during polluted periods. As
shown in Fig. 9, the time series of factor 4 was highly correlated between
ground level and 260 m (r2=0.74), although the peak diameter in size
distributions was slightly different (114 and 98 nm, respectively). These
results suggest a similar source of factor 4 at different altitudes. The
diurnal cycle of factor 4 was also similar at the two different heights,
which both showed a small noon peak and high concentrations at night. Such a
diurnal cycle was similar to that of less oxidized secondary organic aerosol (SOA) observed during
the same study (Zhao et al., 2017). Therefore, we inferred that
factor 4 is a secondary factor that was associated with photochemical
processing and semi-volatile species. Compared to factor 4, factor 5 showed
the best correlation between the two heights (r2=0.91), and the
time series and diurnal cycles were remarkably similar to those of highly
oxidized SOA and sulfate (Zhao et al., 2017), indicating that factor 5 is an aged secondary factor and was mainly formed on a regional scale.
Consistently, the bivariate polar plot of factor 5 showed a dominant source
region to the south, supporting a major influence of regional transport from
the south. Regional emission controls showed large yet different impacts on
factor 4 and factor 5. While the average number concentrations of factor 4
showed decreases of 49 and 37 % at ground level and 260 m,
respectively, during the control period, those of factor 5 had the most
reductions of 65 and 74 %, respectively. These results are consistent
with our previous conclusions that regional emission controls have the most
impacts on highly aged secondary aerosols (Sun et al., 2016a; Zhao et al.,
2017).
Overall, the five factors, which are associated with new particle events,
local primary emissions (e.g., cooking and traffic emissions), and secondary
formation with different aging process, represent the major sources of particle numbers in
the megacity of Beijing. The contribution of secondary
sources was dominant at 260 m throughout the day by varying from
∼ 50 to 80 % (Fig. 10b), and the average contribution
(60 %) was also higher than that (34 %) at ground level. In contrast, the
cooking source was the largest contributor to the total particle numbers, on
average accounting for 33 %. Therefore, our results not only illustrated
the similarities and differences of particle number concentrations and
sources at different altitudes in the megacity, but also demonstrated the
different responses of source factors to regional emission controls.
Average diurnal variations in number fraction of PMF factors at
(a) ground level and (b) 260 m. Panel (c) shows a comparison of the average
diurnal cycles of particle number concentrations for PMF factors at ground
level and 260 m.
Conclusions
We conducted the first simultaneous real-time measurements of particle
number size distribution along with aerosol particle composition at ground
level and 260 m on a meteorological tower in urban Beijing from 22 August to
30 September 2015. Our results showed that the number size distributions
had significant differences between the two heights, although the particle
volume and PM1 mass concentrations were overall similar. The average
number concentration (15–400 nm) was 7473 (±4324) cm-3 at 260 m, which is 26 % lower than that at ground level
(10 134 ± 4680 cm-3). The number concentrations of accumulation particles
(100–400 nm) at 260 m were highly correlated with those at ground level (r2=0.85), indicating their similar sources. However, the correlations were much
weaker for Aitken-mode particles, suggesting that they have more different
sources at different altitudes. A more detailed analysis suggests that the
vertical differences in particle number concentrations varied as functions
of size. While particles in the size range of 100–200 nm showed higher
concentrations at 260 m, those of smaller particles were more dominant at
ground level. These results might indicate the different contributions of
local emissions and regional transport to particle numbers at different
altitudes. We also observed an increase in the ratio of 260 m to ground level for
all particles in different size ranges during daytime, highlighting the
impacts of vertical mixing.
Particle growth events were occasionally observed in this study. The average
particle growth rate was 3.6 nm h-1 at 260 m and 3.2 nm h-1 at
ground level. By comparing with aerosol composition changes
during the growth period, we found that organics appeared to play a more
important role than sulfate during the early stage of the growth (09:00–12:00), while organics and sulfate are both important after that. The
sources of particle numbers were characterized by PMF, and our results
illustrated three common sources at different altitudes, i.e., new particle
formation and growth, local secondary formation, and regional transport. We
also observed much higher primary emissions from cooking sources at ground
level than 260 m, highlighting the importance of local source emissions in
the characterization of NPF and growth events at ground level. In addition, we
found that regional emission controls exerted a large impact on reducing
accumulation-mode particles, for example, by 65–74 % for the regional
factor, while regional emission controls had minor impacts on small Aitken-mode particles, mainly due to
the enhanced NPF events and the limited controls on local source emissions.
These results are overall consistent with the conclusions from our previous
studies during the Asia-Pacific Economic Cooperation (Sun et al., 2016a;
Xu et al., 2015; Chen et al., 2015).