Summertime meteorological fields and air quality in Xi'an
Figure 3a–d show the temporal variations of the temperature, relative
humidity, and wind speed and direction at Xianyang meteorological station
(Fig. 1c) during the summer of 2013. In general, the Guanzhong basin is hot
and humid in the summer, with an average temperature of 26.7 ∘C and
a relative humidity of 67.2 % recorded at the Xianyang station. The winds
are not strong in the basin; the average wind speed is around 3 m s-1
at the Xianyang station. During the simulation period, the observed average
temperature, relative humidity, and wind speed at Xianyang station are
27.9 ∘C, 63.4 %, and 3.4 m s-1, respectively,
representing typical summertime meteorological conditions.
Temporal variations of the observed surface
(a) temperature, (b) relative humidity, (c) wind
speed, and (d) wind direction at Xianyang Meteorological Station and
near-surface (e) O3 and (f) PM2.5 concentrations
averaged over 13 sites in Xi'an during summer of 2013. Red curves depict the
simulation period (22–24 August) in this study.
Monthly minimum, 5th percentile, median, 95th percentile, and
maximum of near-surface O3 concentrations in the afternoon averaged over
13 observational sites in Xi'an from April 2013 to March 2014.
Observed (black dots) and simulated (blue lines) diurnal profiles of
(a) surface temperature and (b) relative humidity averaged
over six meteorological sites from 22 to 24 August 2013.
Observed (black dots) and simulated (blue lines) diurnal profiles of
surface wind speeds at six meteorological sites from 22 to 24 August 2013.
Same as Fig. 6 but for surface wind directions.
Pattern comparison of simulated vs. observed near-surface O3
concentrations at 08:00 and 15:00 CST from 22 to 24 August 2013. Colored
squares: O3 observations; color contour: O3 simulations; black
arrows: simulated surface winds.
The profiles of summertime hourly O3 and PM2.5 concentrations
averaged over 13 sites in Xi'an are also shown in Fig. 3e and f,
respectively, to provide an overview of the air quality in the summer of
2013. The observed average PM2.5 and peak O3 concentrations
frequently exceed 75 and 160 µg m-3, respectively, showing bad
air quality in Xi'an. The simulation period corresponds to a heavy pollution
episode with fairly high O3 and PM2.5 concentrations, which often
occurs during summertime. Figure 4 further presents the monthly minimum,
5th percentile, median, 95th percentile, and maximum observations
of near-surface O3 concentrations in the afternoon averaged over 13
sites in Xi'an during the period from April 2013 to March 2014. The seasonal
cycle of O3 levels in Xi'an shows high summertime O3
concentrations, which is consistent with the observation in the North China Plain (Cooper et
al., 2014). In the study of Cooper et al. (2014), the midday O3 mixing
ratio in the North China Plain peaks in June and then decreases in July and
August due to the southerly monsoon flow. However, during the summer of 2013,
the median O3 concentration in the afternoon in Xi'an increases
progressively from about 90 µg m-3 in June to 120 µg m-3
in August, with the maximum increasing from about 170 µg m-3
in June to 210 µg m-3 in August, which is possibly
caused by the inland location of Xi'an with less monsoon precipitation during
summertime.
Pattern comparison of simulated vs. observed near-surface NO2
concentrations at 08:00 and 15:00 CST from 22 to 24 August 2013. Colored
squares: NO2 observations; color contour: NO2 simulations; black
arrows: simulated surface winds.
Comparison of measured (black dots) and simulated (blue line)
diurnal profiles of near-surface hourly (a) O3 and (b) NO2 averaged
over all ambient monitoring stations from 22 to 24 August 2013.
Panel (a): scatterplot of measured daily aerosol constituents with
simulations and comparison of (b) measured and (c) modeled PM2.5
chemical composition (%).
Table 2 shows the comparison of summertime O3 and PM2.5
concentrations (averaged in the afternoon) in Xi'an to those in the main cities of
BTH, YRD, and PRD in China during 2013. The O3 and PM2.5
concentrations in cities of BTH are much higher than those in Xi'an, showing
the heavy air pollution in BTH. Due to the impact of frequent precipitation
in southern China, the PM2.5 concentrations in the cities of YRD and PRD
are lower than those in Xi'an, but the O3 concentrations in Shanghai
and Hangzhou are still higher than those in Xi'an. Generally, the air
quality in Xi'an is better than that in the cities of BTH but worse than
that in Guangzhou of PRD.
Summertime O3 and PM2.5 concentrations (averaged in the
afternoon) in the main cities of Guanzhong basin, BTH, YRD, and PRD in China
during 2013.
Region
City
O3 (µg m-3)
PM2.5 (µg m-3)
Guanzhong
Xi'an
104.6
48.5
BTH
Beijing
133.9
74.7
Tianjin
116.9
78.1
Shijiazhuang
140.4
86.6
YRD
Shanghai
122.9
47.1
Hangzhou
110.5
35.0
Nanjing
96.6
41.2
PRD
Guangzhou
94.9
29.4
Pattern comparison of simulated vs. observed near-surface PM2.5
concentrations at 08:00 and 15:00 CST from 22 to 24 August 2013. Colored
squares: PM2.5 observations; color contour: PM2.5 simulations;
black arrows: simulated surface winds.
Comparison of measured (black dots) and simulated (blue line)
diurnal profiles of near-surface hourly PM2.5 averaged over all ambient
monitoring stations from 22 to 24 August 2013.
Retrieved (black dots) and calculated (blue lines) diurnal profiles
of (a) AOD and (b) aerosol SSA at 440 nm at the IEECAS
site from 22 to 24 August 2013 and pattern comparison of calculated vs.
retrieved AOD at 550 nm at 10:00 CST on (c) 22 August and
(d) 23 August 2013. Colored squares: retrieved AOD; color contour:
calculated AOD.
Diurnal variations of the change in (a) J[NO2] and
(b) O3 concentrations averaged in Xi'an and surrounding areas due to aerosol
effects from 22 to 24 August 2013.
Diurnal profiles of (a) O3 concentrations and (b) O3
changes averaged in Xi'an and surrounding areas caused by a 50 % reduction
in anthropogenic NOx and VOC emissions, respectively, from 22 to
24 August 2013. Blue line: the reference simulation; red line: the simulation with
a 50 % reduction in anthropogenic NOx emissions; green line: the
simulation with a 50 % reduction in anthropogenic VOC emissions.
Change in O3 concentrations in the bottom model layer, averaged
during O3 peak time from 22 to 24 August 2013 due to a 50 % reduction
in anthropogenic (a) NOx and (b) VOC emissions and the 3-day average
ratio of P(H2O2) / P(HNO3) during O3 peak time.
Diurnal variations of contributions of biogenic emissions to
near-surface isoprene and O3 concentrations averaged in Xi'an and
surrounding areas from 22 to 24 August 2013.
Diurnal profiles of (a) O3 concentrations and (b) O3
contribution from various anthropogenic sources averaged in Xi'an and
surrounding areas from 22 to 24 August 2013. Blue line: the reference
simulation; brown line: the simulation without industry emissions; green
line: the simulation without residential emissions; red line: the simulation
without transportation emissions.
Diurnal profiles of (a) O3 concentrations and
(b) O3 changes averaged in Xi'an and surrounding areas caused
by a 50 % reduction in various anthropogenic sources from 22 to
24 August 2013. Blue line: the reference simulation; brown line: the
simulation with a 50 % reduction in industry emissions; green line: the
simulation with a 50 % reduction in residential emissions; red line: the
simulation with a 50 % reduction in transportation emissions; the black
line: the simulation with a 50 % reduction in all anthropogenic emissions.
Model performance
For the purpose of this discussion, we have defined the reference simulation as that in
which the emissions from various sources and aerosol effects on the
photochemistry are included (hereafter referred to as REF), and results from
the reference simulation are compared with the observations in Xi'an.
Meteorological field simulations
Considering that the meteorological conditions play a crucial role in air
pollution simulations (Bei et al., 2008, 2010, 2012), which determine the
accumulation or dispersion of pollutants, verifications are first performed
for the simulations of meteorological fields in Xi'an and surrounding areas.
Figure 5 shows the temporal profiles of the simulated and observed surface
temperature and relative humidity averaged over six meteorological
observation sites from 22 to 24 August 2013. The WRF-Chem model reproduces
successfully the temporal variations of the surface temperature during the
3-day episode, but in general the model slightly underestimates the
observations, particularly in the morning (Fig. 5a). The MB and RMSE are -0.76
and 1.1 ∘C, respectively, and the IOA reaches 0.97,
indicating good agreement of the surface temperature simulations with
measurements (Table 3). The WRF-Chem model generally tracks the temporal
variations of the surface relative humidity well, with an IOA of 0.92 (Fig. 5b).
However, model underestimation of the observed surface relative
humidity is obvious at midnight and during morning. The MB and RMSE are -4.5 and
5.5 % for the surface relative humidity simulations, respectively.
Statistical comparison of simulated and measured O3, NO2,
PM2.5, temperature, relative humidity, and wind speed at monitoring
sites from 22 to 24 August 2013.
Predictands
Classification
MB
RMSE
IOA
O3 (µg m-3)
Averaged
-9.0
29.0
0.91
NO2 (µg m-3)
Averaged
-5.2
11.0
0.73
PM2.5 (µg m-3)
Averaged
-1.4
21.0
0.92
Temperature (∘C)
Averaged
-0.76
1.1
0.97
Relative humidity (%)
Averaged
-4.5
5.5
0.92
Wind speed (m s-1)
Xi'an
1.7
2.1
0.26
Xianyang
1.3
1.5
0.61
Jinghe
0.14
1.1
0.74
Lintong
1.2
1.5
0.63
Chang'an
0.69
1.2
0.47
Lantian
0.43
1.1
0.61
Figures 6 and 7 present the comparisons of simulated and observed wind
speeds and directions at six meteorological observation sites from 22
to 24 August 2013, respectively. The model fails to yield the observed temporal
variation of the wind speed at the Xi'an site with the IOA of 0.26 and also
significantly overestimates the observation with a MB of 1.7 m s-1 and
RMSE of 2.1 m s-1. In addition, fluctuating wind direction was observed at
the Xi'an site; in contrast, the simulated winds remained fixed in a
northeasterly direction until the evening of 23 August and then changed to the southwest
before noontime of 24 August. As the Xi'an site is located in the urban
center of Xi'an, it is surrounded by high buildings which significantly
alter the airflow on the ground surface, causing light and disordered
surface winds. Although the urban canopy model is utilized in the WRF-Chem
model simulations, the simulated surface winds are still biased considerably
in the urban region due to the simplification of building distributions and
heights and the inability of the model for microscale simulations (Chen et
al., 2011; Lee et al., 2011). The simulated winds at the Xi'an site are similar
to those at the Lintong, Xianyang, and Jinghe sites in the north of Xi'an, but
in general, the wind simulations at these three sites are in good agreement
with the observations, with an IOA of at least 0.61 for the wind speed,
further suggesting the impacts of buildings in the urban region on the wind
simulations. The model performs well in predicting the wind speed and
direction at the Chang'an and Lantian sites in the south of Xi'an. It should be
noted that the model considerably overestimates the observed wind speeds at
all observations sites in the early morning of 23 August and also fails to
track the variation of wind directions over the sites in the southern part
of Xi'an, which causes the biased dispersion of the plume formed during
daytime. In addition, the overestimation of the wind speed is also
remarkable in the afternoon and evening of 24 August, and this evacuates the
plume formed in the urban region more efficiently.
Gas-phase species simulations
The modeled O3 and NO2 mass concentrations are compared to the
measurements at the ambient monitoring stations released by China MEP.
Figure 8 shows the spatial distributions of calculated and observed
near-surface O3 concentrations at 08:00 and 15:00 China Standard Time
(CST) from 22 to 24 August 2013, along with the simulated wind
fields. When the northeast wind is prevalent in the Guanzhong basin, due to
the impact of the topography (Fig. 1), stagnant conditions are frequently
formed in Xi'an and surrounding areas. At 08:00 CST, the model
underestimates the observed near-surface O3 concentrations in the urban
area of Xi'an, which is perhaps caused by the titration of NO emitted from
traffic during rush hours. At 15:00 CST, the calculated near-surface O3
distributions are generally consistent with the observations at the ambient
monitoring sites. On 22 August the northeast winds were strong and the
stagnant region with high O3 plume was located in the southwest of Xi'an
and surrounding areas; the predicted O3 concentrations are less than
160 µg m-3 in the urban area of Xi'an, in good agreement with
the measurements. On 23 August the convergence in the urban area of Xi'an was
favorable for the accumulation of O3 precursors and a high O3 plume
was formed in the afternoon, with the O3 concentration exceeding
200 µg m-3. On 24 August the plume formed in the urban region
of Xi'an was pushed to the south of Xi'an and surrounding areas in the
afternoon and the simulated O3 concentrations were less than
200 µg m-3 in the urban area of Xi'an, generally consistent
with the observations. In addition, the O3 level in Xi'an and
surrounding areas was also affected by the O3 transport from its upwind
region. It is worth noting that the model cannot replicate reasonably the
spatial variation of the observed O3 concentration at monitoring sites
in the urban area of Xi'an. Although 3 km horizontal resolution is used in
the study, it still cannot resolve adequately the spatial variation of
O3 concentrations over monitoring sites with a distance of less than
21 km (Skamarock, 2004). Unfortunately, the ambient monitoring sites in
Xi'an are mainly concentrated in the urban area (around
20 km × 30 km); thus, future model studies with higher horizontal
resolution will be needed to improve the near-surface O3 simulations
under the present monitoring site distribution.
The simulated near-surface NO2 distributions agree well with the
observations at the ambient monitoring sites in the morning (Fig. 9a, c,
and e), with the highest NO2 concentration in the urban center of
Xi'an. In the afternoon, the model overestimates the observation in the
urban center on 23 August, while it underestimates the observation on 24 August.
Figure 10 displays the diurnal profiles of predicted and observed
near-surface O3 and NO2 concentrations averaged over the ambient
monitoring sites during the episode. The model tracks the temporal
variations of surface O3 concentrations well during daytime (Fig. 10a),
but the simulated O3 concentrations deviate markedly from the
observations in the early morning hours on 23 August. Apparently, the plume
with high O3 concentrations formed in the southwest of Xi'an during
daytime (Fig. 8b) on 22 August was transported back to the urban area of
Xi'an in the early morning on 23 August causing the observed high O3
level. However, due to biases of the wind simulations in the early morning on
23 August (Figs. 6 and 7), the plume formed from the previous day was not
transported back to the urban area of Xi'an, leading to the remarkable
underestimation of the observed O3 concentrations. The MB, RMSE, and
IOA of the simulated O3 concentration averaged over monitoring stations
are -9.0 and 29 µg m-3 and 0.91, respectively. Although the
model reasonably well reproduces the variation of the observed NO2
concentrations (Fig. 10b), with an IOA of 0.73, it often overestimates or
underestimates the observation. Uncertainties in the simulated meteorological
fields or the emission inventory might be responsible for the model biases in
simulating NO2 distributions and variations.
In summary, the calculated distribution and variation of near-surface
O3 and NO2 concentrations are in good agreement with the
corresponding observations, suggesting that the model simulates well the
meteorological fields and that the emission inventory used in the study is also
reasonable.
Aerosol simulations
Atmospheric particulate matter or aerosols scatter or absorb a fraction of
solar radiation and increase or decrease the photolysis rates in the
atmosphere, influencing the O3 formation. Therefore, in order to
reasonably verify the aerosol impact on the photolysis and O3 level, it
is important to evaluate the simulated aerosol composition, variation, and
distribution using available measurements. Daily measurement of aerosol
constituents is performed using the filter sample at the IEECAS site,
including sulfate, nitrate, ammonium, organic, and elemental carbon.
Figure 11a presents a scatterplot of the measured versus calculated daily
mean concentration of aerosol constituents at the IEECAS site from 22 to
24 August 2013. It should be noted that the simulated organic aerosol is
compared with the filter-measured organic carbon scaled by a factor of 2
(Carlton et al., 2010). The model performs well in simulating daily mean
sulfate and organic aerosol concentrations. The model tends to overestimate
the observed nitrate and ammonium concentration; this might be caused by the
nitrate loss due to the evaporation from filters in the summer (Ianniello et
al., 2011). The simulated daily mean elemental carbon concentrations deviate
from the measurements considerably on 22 and 23 August, which is perhaps
caused by the daily variations of elemental carbon emissions. Comparison of
the observed and modeled PM2.5 mass composition averaged during the
3-day episode is displayed in Fig. 11b and c. Sulfate is the dominant
constituent of the observed PM2.5 at the IEECAS site, consisting of
around 39 % of the PM2.5 mass, and the simulated sulfate
contribution to the PM2.5 mass is about 35 % on average, close to
the observation. The high sulfate concentrations come mainly from the
SO2 heterogeneous reaction on aerosol surfaces under humid conditions
(Wang et al., 2014). The measured and modeled organic aerosols make up about
19 % of the PM2.5 mass at the IEECAS site, and secondary organic
aerosol contributes more than 50 % of the modeled organic aerosol due to
the high atmospheric oxidation capacity in the summer. The modeled ammonium,
nitrate, and elemental carbon account for about 15, 6.8, and 3.1 % of the
PM2.5 mass, respectively, comparable to the observed 14, 6.6, and
3.6 % at the IEECAS site.
Figure 12 shows the simulated geographic distributions of near-surface
PM2.5 mass concentrations and the observations over the monitoring
stations at 08:00 and 15:00 CST from 22 to 24 August. On 22 August the
convergence is formed in the north of Xi'an and surrounding areas at
08:00 CST, leading to the buildup of pollutants and high PM2.5
concentrations. The simulated PM2.5 concentrations are more than
75 µg m-3 in the north of Xi'an and surrounding areas,
consistent with the measurements, but exceed 150 µg m-3 in
the south, where stagnant conditions are formed, and are much higher than the
observation. At 15:00 CST, well organized northeast winds push the
convergence zone to the southeast of Xi'an and surrounding areas and the
simulated PM2.5 concentrations are less than 75 µg m-3,
in good agreement with the measurements. On 23 August the convergence is held
within in the urban area of Xi'an, causing heavy PM2.5 pollution. The
calculated PM2.5 concentration is more than 115 µg m-3
at 08:00 and 15:00 CST, comparable to the measurements. The observed and
simulated PM2.5 patterns on 23 August are similar to those on 23 August
but the PM2.5 concentrations on 24 August are enhanced. For example, the
observed and simulated PM2.5 concentrations over all monitoring stations
exceed 150 µg m-3 at 08:00 CST. In addition, at 15:00 CST,
due to the overestimation of the wind speed (Fig. 6), the convergence zone is
pushed to the south of Xi'an and surrounding areas, causing the
underestimation of PM2.5 concentration in the north. The model
reproduces the observed diurnal profile of the PM2.5 concentration
averaged over the monitoring stations during the episode (Fig. 13), with a
MB of -1.4 µg m-3, an RMSE of
21 µg m-3, and an IOA of 0.92. Apparently, the convergence
zone location, which is determined by the meteorological fields,
significantly influences the PM2.5 simulations. When the simulated
convergence zone is formed in the south of Xi'an and surrounding areas on the
morning of 22 August, the model considerably overestimates the observed
PM2.5 concentration. However, when the simulated winds are too strong
and the convergence zone is pushed to the south of Xi'an and surrounding
areas on 23 August, the model notably underestimates the observations.
The simulated column-integrated aerosol optical depth (AOD) and single-scattering albedo (SSA) are verified using the available measurements from
the surface site and satellite. The simulated aerosol optical properties are
calculated using the method developed by Li et al. (2011b). Figure 14a and
b show the comparison of simulated column-integrated AOD and aerosol SSA, respectively, at
440 nm with measurements at the IEECAS site, which are retrieved
from the observations of a sun–sky radiometer (Su et al., 2014). The
simulated AOD on 22 August is comparable to the measurements at the IEECAS site,
but the WRF-Chem model overestimates the observation in the morning due to
the overestimation of PM2.5 concentrations. The simulated SSA on 22 August
ranges from 0.92 to 0.95, close to the measurement. However, on 23 August,
the retrieved AOD and SSA exceed 3.0 and 0.95, respectively, higher than the
corresponding simulations. The underestimation of the simulated AOD and
aerosol SSA could be attributed to the underestimation of the relative
humidity, not limited to the ground surface as shown in Fig. 5. The
distribution of the calculated AOD at 550 nm from the MODIS (Moderate
Resolution Imaging Spectroradiometer) aerosol level-2 product at
5 × 5 1 km pixel resolution is also compared with the model results
(Fig. 14c and d). The simulated AOD pattern on 22 August agrees well with the
measurements, except over the Qin Ling where the convection is active. The
model underestimates the observed AOD in the north of Xi'an on 23 August,
which is likely caused by the bias of the simulated relative humidity.
Apparently, the AOD at 550 nm in Xi'an and surrounding areas is high,
exceeding 0.8 on 22 August and 1.0 on 23 August.
Sensitivity studies
Effects of aerosol on the O3 formation
O3 formation in the atmosphere is a complicated photochemical process,
which is determined by its precursors from various sources and
transformation in the presence of sunlight. High AOD in Xi'an and the surrounding areas efficiently scatter or absorbs sunlight to decrease the
photolysis frequencies in the planetary boundary layer (PBL) and further the
O3 formation. High O3 levels enhance atmospheric oxidation
capability and the secondary aerosol formation, increasing the aerosol
concentration in the atmosphere, but conversely, high aerosol levels
decrease the photolysis frequencies and suppress the O3 formation in
the PBL. The interactions of O3 with aerosols complicate the design of
O3 control strategies.
The aerosol effect on O3 formation in Xi'an and surrounding areas is
examined by the sensitivity study without aerosol effects on the photolysis
compared to the reference simulation (hereafter, we define the sensitivity
simulation as SEN). Figure 15a and b present the diurnal profiles of the
change in the NO2 photolysis frequency (J[NO2]) and O3
concentration, respectively, averaged in Xi'an and surrounding areas due to aerosol effects
from 22 to 24 August. Aerosols significantly decrease
J[NO2] by 30–70 % (defined as (REF-SEN)/SEN) in the early morning
and late afternoon hours when the solar zenith angle is large, showing the
impact of a long aerosol optical path for incoming radiation. Due to the high
aerosol level, the aerosol effect on J[NO2] is still substantial
during noontime, decreasing J[NO2] by over 20 % on 23 and 24 August
when the plume is stagnant in the urban region of Xi'an. The aerosol effect
on the photolysis frequency in this study is larger than those reported in
other studies (e.g., Jacobson, 1998; Li et al., 2005, 2011b). The aerosol
impact on O3 formation is most significant during the late morning and
early afternoon (Fig. 15b). On average, in Xi'an and surrounding areas, the
reduction in O3 concentration (defined as (REF-SEN)) due to the aerosol
effect on photolysis is less than 20 µg m-3 on 22 August but
more than 30 µg m-3 during noontime on 23 August and over
50 µg m-3 in the late morning on 24 August. The aerosol effect on
O3 formation in this study is comparable to those reported by Castro et
al. (2001) in Mexico City. It should be noted that the impact of photolysis
on O3 level varies depending on the ratio of VOCs to NOx (Stockwell
and Goliff, 2004). The important aerosol effects on O3 formation may
pose a dilemma for O3 control strategies. If O3 concentrations are
decreased by reducing its precursor's emissions, the aerosol level will also
decrease due to direct and indirect contributions from the emission control,
which compensates for the O3 reduction by enhancing the photolysis
frequency.
O3 response to emission changes
In the urban area, when the meteorological conditions are favorable for the
accumulation of pollutants in the PBL, the O3 precursors of
anthropogenically or naturally emitted VOCs and NOx, react chemically in
the presence of sunlight, leading to high O3 level. In order to devise
an effective O3 control strategy, it is important to investigate the
regime of O3 production. The regime of O3 production in Xi'an and
surrounding areas is examined using sensitivity studies by reducing
anthropogenic VOC (AVOC) or NOx emissions by 50 % in the WRF-Chem
simulations. Figure 16 compares the near-surface O3 concentrations
averaged in the urban area of Xi'an in the reference simulation to the two
sensitivity studies in which AVOCs and NOx are decreased by 50 %. A
50 % reduction in AVOCs decreases the O3 concentration averaged in
the areas around Xi'an consistently during the episode, particularly during
peak time (defined as 14:00–16:00 CST hereafter). A 50 % reduction in NOx enhances
the O3 level in the morning due to the decrease in NO emission; but in
the afternoon, it decreases the O3 level, similar to the effect from a
50 % reduction in AVOCs, leading to a complicated O3 production
regime.
Figure 17a shows the 3-day average near-surface O3 change during peak
time with a 50 % reduction in NOx emissions (defined as O3(SEN)
– O3(REF)). In Xi'an and surrounding areas, except the urban center,
the simulated average O3 concentrations are decreased by about
10–40 µg m-3 due to a 50 % reduction in NOx. In the
urban center, the O3 concentrations are slightly enhanced (less than 10 µg m-3) with the high decrease in NOx emissions. A 50 % reduction
in AVOC emissions consistently reduces the O3 concentration in Xi'an and
surrounding areas by up to 40 µg m-3 (Fig. 17b). The response
of O3 change to a 50 % reduction in NOx or AVOC emissions
cannot clearly indicate the O3 production regime in Xi'an and
surrounding areas.
Sillman (1995) proposed that the ratio of the production rates of hydrogen
peroxide and nitric acid (P(H2O2) / P(HNO3)) can be
used to investigate the sensitivity of ozone formation to the precursors. If
the ratio of P(H2O2) / P(HNO3) is less than 0.3, the
O3 production regime is VOC-sensitive. If the ratio exceeds 0.5, the
regime is NOx-sensitive. The ratio ranging from 0.3 to 0.5 indicates the
transition from NOx- to VOC-sensitive regime. Figure 17c displays the
distribution of the 3-day average P(H2O2) / P(HNO3)
during the O3 peak time. In Xi'an and surrounding areas, the
P(H2O2) / P(HNO3) ratio varies from 0.2 to 1.0,
suggesting that the O3 production regime is very complicated. In the
south of Xi'an and surrounding areas, the O3 production regime lies in
the transition from NOx to VOC-sensitive chemistry. The analyses using
the P(H2O2) / P(HNO3) indicator and the results
obtained from the two sensitivity studies suggested that it is not
straightforward to devise effective O3 control strategies for Xi'an and
the surrounding areas.
O3 contribution from natural and anthropogenic sources
Biogenic emissions provide natural O3 precursors and numerous studies
have shown that biogenic VOCs play an important role in ground-level O3
formation in the urban areas (e.g., Solmon et al., 2004; Li et al., 2007),
thus complicating O3 control strategy. A sensitivity study without
biogenic emissions is conducted and compared with the reference simulation to
evaluate the contribution of biogenic emissions to ozone production. During
noontime, biogenic emissions contribute about 0.3 ppb isoprene averaged in
Xi'an and surrounding areas, and the O3 contribution from biogenic
emissions is around 10 µg m-3 (Fig. 18). Large amounts of
biogenic emissions are released over the Qin Ling to the south of Xi'an, and
can be transported to the urban area under favorable meteorological
conditions, enhancing O3 formation. However, the northeast wind is
prevalent in the Guanzhong basin during daytime, which is not favorable for
the transport of biogenic emissions from the Qin Ling. Although the O3
level enhanced by biogenic emissions is not significant in Xi'an and
surrounding areas, the high reactivity of biogenic VOCs, such as isoprene and
monoterpenes, will play an increasingly important role in O3 formation
when the anthropogenic VOCs are decreased as a result of O3 control
measures.
We have further used the sensitivity studies to evaluate the contribution of
anthropogenic emissions from industrial, residential, and transportation
sources to O3 production. The industrial emissions contribute more than
70 % of the anthropogenic VOCs and play the most important role in the
O3 formation in Xi'an and surrounding areas during daytime, compared to
residential and transportation emissions. On average, the near-surface
O3 contribution from industrial emissions is about
10–30 µg m3 in the afternoon and exceeds
20 µg m-3 during O3 peak time (Fig. 19b). Transportation
emissions contribute about 10 to 20 µg m-3 ozone in the
afternoon, while residential emissions contribute less than
10 µg m-3 O3.
Occurrence days of the defined PM2.5 and O3 exceedance
levels during the summer of 2013.
Beijing
Tianjin
Shijiazhuang
Jinan
Taiyuan
Xi'an
1Level I
57
65
64
72
53
61
2Level II
33
41
43
41
28
20
1 Hourly PM2.5 and O3 concentrations exceeding 35 and
160 µg m-3, respectively.
2 Hourly PM2.5 and O3 concentrations exceeding 75 and 200 µg m-3, respectively.
The original data are from China MEP.
Sensitivity studies have shown that there is no single anthropogenic ozone
precursor emission source that dominates the O3 level in Xi'an and
surrounding areas. The simulation without the most important industrial
source still predicts high near-surface O3 concentrations in Xi'an and
surrounding areas (Fig. 19a). The O3 production regime in Xi'an and
surrounding areas varies from NOx- to VOC-sensitive chemistry,
constituting one of the possible reasons for the insensitivity of O3
concentration to the emission change. Additionally, in the case of high
aerosol levels, aerosol effects on photolysis also compensate for the O3
decrease through enhancing photolysis frequencies due to a decrease in
aerosol concentrations caused by the emission reduction. Although biogenic
emission does not play a major role in the O3 formation in Xi'an and
surrounding areas, it provides reactive a VOC precursor for O3
formation. Therefore, with high O3 and PM2.5 in Xi'an and
surrounding areas, decreasing emissions from various anthropogenic sources
alone cannot efficiently mitigate the O3 pollution. Sensitivity studies
have been performed to further demonstrate the difficulties in devising
O3 control strategies through decreasing anthropogenic emissions from
industry, residential, transportation, and all anthropogenic sources by
50 % in the WRF-Chem simulations. A 50 % reduction in industrial
emissions only resulted in less than a 7 % decrease in O3
concentrations in Xi'an and surrounding areas (Fig. 20). Even if all the
anthropogenic emissions are reduced by 50 %, the decrease in O3
concentrations is not more than 14 %.