Environmental Pollution 239 (2018) 818e828

Contents lists available at ScienceDirect

Environmental Pollution
journal homepage: www.elsevier.com/locate/envpol

A proposed methodology for impact assessment of air quality traffic-

related measures: The case of PM2.5 in Beijing*
^nia Fontes a, b, *, 1, Peilin Li a, c, Nelson Barros b, Pengjun Zhao a
Ta
a
Centre for Urban Planning and Transport Studies, Peking University, Beijing, China
b
FP-ENAS - UFP Energy, Environment and Health Research Unit of University Fernando Pessoa, Porto, Portugal
c
Transport Studies Unit, University of Oxford, Oxford, OX2 7TL, UK

a r t i c l e i n f o a b s t r a c t

Article history: Air quality traffic-related measures have been implemented worldwide to control the pollution levels of
Received 25 October 2016 urban areas. Although some of those measures are claiming environmental improvements, few studies
Received in revised form have checked their real impact. In fact, quantitative estimates are often focused on reducing emissions,
11 April 2018
rather than on evaluating the actual measures’ effect on air quality. Even when air quality studies are
Accepted 13 April 2018
Available online 9 May 2018
conducted, results are frequently unclear.
In order to properly assess the real impact on air quality of traffic-related measures, a statistical
method is proposed. The method compares the pollutant concentration levels observed after the
Keywords:
Statistical method
implementation of a measure with the concentration values of the previous year. Short- and long-term
Traffic measure impact is assessed considering not only their influence on the average pollutant concentration, but also
Pollution on its maximum level. To control the effect of the main confounding factors, only the days with similar
Air quality environmental conditions are analysed. The changeability of the key meteorological variables that affect
Meteorology the transport and dispersion of the pollutant studied are used to identify and group the days categorized
as similar. Resemblance of the pollutants' concentration of the previous day is also taken into account.
The impact of the road traffic measures on the air pollutants’ concentration is then checked for those
similar days using specific statistical functions.
To evaluate the proposed method, the impact on PM2.5 concentrations of two air quality traffic-related
measures (M1 and M2) implemented in the city of Beijing are taken into consideration: M1 was
implemented in 2009, restricting the circulation of yellow-labelled vehicles, while M2 was implemented
in 2014, restricting the circulation of heavy-duty vehicles. To compare the results of each measure, a
time-period when these measures were not applied is used as case-control.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction emissions, has led to a record-high number of heavy haze pollution

In the recent years, Beijing's population has been growing
quickly. Between 1990 and 2015, it increased from 10.8 to 21.7
million people and, today, the city is one of the most populated
cities in the world (BJSTATS, 2016). This fast growth, combined with
stable atmospheric conditions and with the transport of regional
*
This paper has been recommended for acceptance by Eddy Y. Zeng.
* Corresponding author. INESC TEC e INESC Technology and Science, Campus da
Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, 4200-465
Porto, Portugal.
E-mail address: tania.d.foontes@inesctec.pt (T. Fontes).
1
Current address: INESC TEC e INESC Technology and Science and FEUP- Faculty
of Engineering, University of Porto, Porto, Portugal.
days during the last years (Cai et al., 2017; Ding and Liu, 2014;
Zhuang et al., 2014). As result, the high levels of PM2.5 concentra-
tions (particulate matter with particle aerodynamic diameter
smaller than 2.5 mm) in the atmosphere have been pointed out as
the main reason for the poor air quality in the city (Batterman et al.,
2016; Guo et al., 2017).
Several PM2.5 episodes observed in Beijing have been attracting
the attention of researchers worldwide (e.g. Fontes et al., 2017;
Huang et al., 2014; Zhao et al., 2011; Wang et al., 2013). Besides the
impact on the climate (Maricp, 2013), this pollutant has also an
important impact on human health. PM2.5 tend to penetrate into
the gas exchange regions of the lung and blood streams causing
permanent DNA mutations, cardiovascular diseases and premature
deaths (Roy et al., 2012). Several studies conducted in China

https://doi.org/10.1016/j.envpol.2018.04.061
0269-7491/© 2018 Elsevier Ltd. All rights reserved.
 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828
reported acute (Chen et al., 2013; Li et al., 2013) and chronic (Chu
et al., 2015) human health problems. A recent study involving
312,944 people in nine European countries revealed that, for every
increase of 5 mg.m
3 in PM2.5, lung cancer rate rose by 18%
(Raaschou-Nielsen et al., 2013). Although research studies revealed
that no safe level of PM2.5 concentrations can be defined to mitigate
the impact on human health, the World Health Organization (WHO,
2005) proposed guidelines for this pollutant by defining an annual
(10 mg.m
3) and a daily (25 mg.m
3) exposure limit. Three interim
targets (IT) were also defined for PM2.5 that must be achievable by
each country with successive and sustained abatement measures.
The most IT permissive levels were adopted by countries as China:
35 mg.m
3 for the annual exposure limit and 75 mg.m
3 for the daily
exposure limit.
To understand the problems of PM2.5, several source appor-
tionment studies have been conducted in Beijing (e.g. Li et al., 2015;
Yang et al., 2016). In a recent study, Li et al. (2015) explain that
40e60% of total PM2.5 emissions in the city are from local emis-
sions. Furthermore, it was also seen that road traffic emissions
contribute between 10 and 20% of the overall PM2.5 emissions of the
city (Huang et al., 2014; Li et al., 2015; Yang et al., 2016). For the
greater Beijing area, the contribution of local emissions can be even
higher (Li et al., 2015). According to the literature, the hotspots of
PM2.5 concentrations in the cities are usually located near road
traffic lanes (Huang et al., 2012) and such levels decrease when the
distance to the main roads increases (Kimbrough et al., 2013; Barros
et al., 2013). All the above together suggests that local controls
deserve high priority and that the management of the road traffic
sector may have here an important role.
Beijing is one of the cities with the highest private car owner-
ship (Shen et al., 2014). In 2030, the number of vehicles is expected
to be 10.4 million (Wu et al., 2011). Both commuting distance and
time put the city in the top of Chinese cities with highest levels of
traffic congestion. In 2014, the daily average commuting distance
was 19.2 km, while the national average was 9.2 km (Chen, 2015).
On top of that, the average commuting time has been rising
continuously: from 38 min in 2005 to 45 min in 2010, and 97 min in
2014 (Xu et al., 2011; Lai, 2015).
In Beijing, to minimise the impact of road traffic, high in-
vestments have been done in public transport. Having exceeded
500 km of lines in 2015, the effort to build a metro network with
1000 km of lines until 2020, one of the longest metro networks in
the world (Yanan, 2015), does not seem to relieve the burden
related to high PM2.5 concentrations in the atmosphere of the city.
Such investments are not new. Over the last 25 years, the Beijing
government has implemented many vehicle control strategies in
order to minimise the effects of the fast growth (Wu et al., 2011;
Zhou et al., 2010). Figure S1 (in the Supplements section) shows an
overview of the main air quality traffic-related measures imple-
mented between 1990 and 2015.
The air quality traffic-related measures introduced in Beijing
during the last few decades have been mostly focused on control-
ling the total number of vehicles and the quality of their engines. As
a result, measures for new and in-use vehicles have been defined. In
1994, a study conducted by the State Environmental Protection
Administration of Beijing explained that more stringent emission
standards were required and, in 1997, a dramatic increase in new
light-duty vehicles was underway (Wu et al., 2011). As a result,
Beijing started to implement emission standards in 1999 in an
accelerated mode and Euro 6 emission standards are expected to be
implemented until the end of 2023 (DieselNet, 2018).
As for in-use vehicles, in 1998, an environmental labelling
measure was implemented. The vehicles registered in Beijing were
issued with a yellow or green label indicating their emission
standard. Since 2003, the yellow-labelled vehicles (also known as
819
high-emitting vehicles) are not allowed to circulate in the Second-
Ring Road (z60 km2, inside the urban area) and later this restric-
tion was expanded to the Sixth-Ring Road (z2000 km2, inside the
urban area) (BENDIBAO, 2013). In 2014, the yellow-labelled vehicles
registered outside of Beijing were also forbidden to run within
Beijing administrative area (BJLZJ, 2014).
Similar measures to the labelling policy were implemented in
Beijing for specific vehicle categories, namely motorcycles and
heavy trucks. Motorcycles were first banned in 2001 and freight
vehicles in 2004 (during daytime 06:00 a.m.e23.00 p.m.). Both
vehicle categories were initially bann from driving within the
Fourth-Ring Road (z300 km2, inside the urban area), but later
these restrictions were enlarged for some of the classes of these
categories until the Sixth-Ring Road (z2000 km2, inside the urban
area) and Fifth-Ring Road (z700 km2, inside the urban area),
respectively (BJJTGL, 2009; BJLZJ, 2014; BJGOV, 2014). For freight
vehicles, the restrictions were even higher, since the number of
allowed hours to circulate also decreased. Such restrictions were
applied for all vehicles; among them, non-Beijing vehicles were
severely restricted both spatially and temporally. In spite of all the
efforts, these measures were not enough to reduce traffic conges-
tion and control the air pollution. Therefore, to remove the oldest
and poorly maintained vehicles available in the city, car-scrapping
incentives were introduced in 2009 (Minghao, 2011).
Fuel quality and alternative fuel type's policies were also
implemented in Beijing city. Restrictions on the lead, sulphur and
benzene contents on gasoline were some of those policies. After
that, the introduction of alternative fuel vehicles and advanced
vehicles brought a new hope to the city. The use of natural gas in
the public bus fleet was being promoted since 1999 and, since 2009,
hybrid electrical-diesel buses started to be used in Beijing. Retrofit
programs to convert gasoline taxis to Flexible-Fuelled Vehicles
(FFV) with either gasoline or Liquefied Petroleum Gas (LPG) were
implemented as well.
Recently, advanced vehicles, also known in China as new-energy
vehicles, are the main focus of the air quality traffic-related mea-
sures. The 2008 Olympic Games was used to launch this policy.
During the event, more than 50 hybrid electric vehicles, pure
electric vehicles and fuel cell vehicles were used (Wu et al., 2011).
Also, several temporary measures were implemented during this
time (e.g. Wang et al., 2010b). Private cars could only operate on
odd or even days based on the last digit of their license plate and
trucks could only operate inside the Sixth-Ring Road at night (12
a.m.e06 a.m.). A demo program restrict drivers on one day a week
depending on what number the license plate ends with. This last
measure, implemented in 2008, was extended until 2011 with huge
social protest and economic losses. Nonetheless, other similar
measures additionally were applied (Zhou et al., 2010; Wu et al.,
2011).
With the Clean Air Act Plan for 2013e2017, more stringent
emission controls were defined. The implementation of Euro 6
emission standards and the improvement of the fleet of public
buses and taxis were two of the main goals of this plan. The gov-
ernment also plans to phase out all yellow-labelled vehicles, setting
a target of no more than 6 million vehicles at the end of 2017 (Zhao,
2015; Zhang et al., 2014).
The environmental impact of traffic restriction measures
implemented in Beijing, namely on PM2.5 emissions, has been
studied by several researchers. Hao et al. (2006) concluded that,
despite the rapid increase in the number of vehicles, total vehicular
emissions have not increased, and traffic measures as the retrofit
programs showed little emission benefits. More recently, Zhang
et al. (2014) reported two decades trend of vehicle emissions in
the city. The results show that traffic control is playing an essential
role in mitigating urban vehicle emissions. This is particularly
820 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828
important for PM2.5 emissions’ control. Shen et al. (2014) found a
significant decline of PM2.5 road emission when vehicle technology
was improved. Still, the results of these studies are not reflected in
the emission studies conducted recently (e.g. Wang et al., 2010a;
Huo et al., 2012). Other work studied the PM2.5 emission trends in
worldwide areas outside China. In Europe, PM2.5 emissions show a
decline since 2010/11 due to the use of diesel particle filters
(Harrison and Beddows, 2017).
At air quality level, temporary measures implemented in Beijing
during the 2008 Olympic Games were also assessed (Figure S1 e
Supplements section). During the event, traffic flow was reduced by
22 %e43%, when compared to the non-traffic control periods (Zhou
et al., 2010; Wang and Xie, 2010). Nevertheless, the impact
assessment of the temporary measures on PM2.5 concentrations is
not consistent among the authors. While some of them suggest no
significant changes (Wang et al., 2010b; Okuda et al., 2011), others
advocate some improvements (Chen et al., 2014).
Different methods have been used to assess the impact of traffic
measures on air quality. Short-term (Wang et al., 2010b) and long-
term (Chen et al., 2014; Okuda et al., 2011) analyses have been
conducted using mostly passive diffusive techniques for PM2.5 data
collection (e.g. Chen et al., 2014; Wang et al., 2010b; Okuda et al.,
2011). Some of the studies compare concentrations with a refer-
ence period (Okuda et al., 2011) or apply complex methods such as
principle component analysis (Chen et al., 2014). The influence of
meteorological effects has been studied using multilinear regres-
sion models (Ordo n
~ ez et al., 2005), generalized additive models and
backward trajectories (Wang et al., 2010b). Although these models
were used to analyse the influence of meteorology, the similarity of
meteorological conditions between distinct time-periods was not
discussed in these works. This is very important to assess the
impact of urban policies, since the effects of day-to-day variations
in meteorology on pollutant concentrations often mask more subtle
effects of the measures (Holman et al., 2015), preventing a clear
understanding of their real impact. According to Cyrys et al. (2014),
an adequate adjustment of the meteorological conditions can only
be made over long periods, preferably one year or more.
The literature review allowed us to identify many air quality
traffic-related measures implemented in Beijing; nevertheless, the
environmental impact claimed by some of those traffic measures
was not fully investigated. Although the long-term impact on
emissions was analysed by some researchers, the effects on air
quality stayed a poorly studied subject. To the best of our knowl-
edge, in Beijing, only the impact of some temporary air quality
traffic-related measures was analysed, and such conclusions were
not consistent among authors (Wang et al., 2010b; Okuda et al.,
2011; Chen et al., 2014). Data collected through passive diffusive
techniques was used in some of these cases (e.g. Chen et al., 2014).
In fact, the results obtained with this measurement technique are
very limited. Sampling rates of days or even weeks are not able to
characterise the fast dynamics of the atmosphere. Furthermore,
some studies reveal that passive diffusive techniques have impor-
tant limitations, since, during the sampling collection, significant
loss of semi-volatile species might happen due to evaporative
processes (Dong et al., 2012). Moreover, the effects of meteorology
often compromise the measures effect.
Taking into account the above limitations, the main goal of this
paper is to propose a general methodology able to assess the
effective impact of traffic-related measures on the concentration
levels of air pollutants in an urban area. To evaluate such meth-
odology, the impact of two air quality traffic-related measures
implemented in Beijing on the concentration levels of PM2.5 is
analysed.
2. Material and methods
In this work, a methodology based on a statistical approach is
proposed in order to assess the impact of traffic-related measures
on air quality. The methodology includes the comparison of the air
quality values observed before and after the implementation of a
given air quality traffic-related measure considering the influence
of the main confounding factors. Two traffic measures imple-
mented in the city of Beijing are selected to demonstrate its
applicability. The next sections present a detailed description of the
methodology proposed as well as the data collected to assess it.
2.1. Method
There are several factors playing an important role on atmo-
spheric transport and dispersion of air pollution. The meteorolog-
ical conditions and the emissions are usually classified as the main
ones (Seinfeld and Pandis, 2016). Thus, to develop a robust
assessment methodology, the influence of meteorological condi-
tions and background concentrations, as well as the effect of other
measures, should be taken into account. In the proposed method,
some criteria were established in order to include these factors.
First, due to the high influence of meteorological conditions on
transport and pollutant dispersion, only similar meteorological
days are compared. Second, to discard the effect of background
concentrations, only air quality measurements with similar con-
centration levels in the previous day to the day under analysis are
selected. Last, the traffic measure to be evaluated must be imple-
mented in isolation. When this is not possible, several factors must
be taken into account in their assessment:
(i) Traffic measure with a potential impact on the pollutant
under study must start in a different time-period than the
measures under evaluation. This allows not only to study the
sole impact of that measure, but also the impact caused by
the synergy with other measures;
(ii) The impact of traffic measures on different pollutants de-
pends of its nature. This means that some measures have no
impact on some pollutants. This is the case of the impact on
PM2.5 concentrations of the restriction on the limit of ben-
zene content in gasoline;
(iii) No significant impact on the control of air pollution is ex-
pected from traffic measures with low coverage. This is the
case of car-scraping or the demo program to introduce
advanced vehicles in Beijing during 2008 (Wu et al., 2011);
(iv) For temporary traffic control measures, a return to the initial
conditions is expected after the resolution of those measures
(e.g. traffic restrictions introduced during the Olympic
Games in Beijing in 2008).
Following the established requirements, the proposed meth-
odology comprises of three steps. First, key meteorological pre-
dictors are selected for the pollutant under evaluation. Second,
based on this information, the days before and after the imple-
mentation of each measure are compared and the days with similar
environmental conditions are identified. This allows the influence
from the variability of the main environmental external factors to
be excluded. Finally, levels of air concentrations for the ‘similar
days’ are assessed to measure the impact of the traffic measure. This
approach is distinct from the ones found in literature for three main
reasons: (i) the impact of traffic measures is assessed on air quality
instead of on emissions; (ii) detailed data from a monitoring station
is used in order to capture the meteorological variabilities observed
through the day; and (iii) besides a meteorological analysis, a set of
rules allow to obtain datasets with similar environmental
 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828 821

conditions between them. datasets have an IOA value equal to 1.0. The FB (eq. (4)) is a statis-
The evaluation impact of each traffic measure covers a period of tical measure recommended by USEPA (1992) to compare air
two years: the year after the implementation of a measure (Vy) is quality datasets. This measure is based on a linear scale and such
compared with the previous one (Vy-1). Such period is defined in values vary between þ2, which indicates an increase of the values
order to minimise the influence of other measures. As the first comparing with the reference value, and
2, which indicates a
months are usually considered as a transitional period for the decrease of the values comparing with the reference value. Equal
adaptation of drivers to a new measure, in this work, the first three datasets have an FB equal zero. NMSE (eq. (5)) is an estimator of the
months are used to assess the short-term impact. The remaining overall deviations between two datasets. While this measure re-
period is used to measure the long-term impact. Fig. 1 shows a flects both systematic and non-systematic deviations, the FB as-
hypothetical example of this evaluation process. The results ob- sesses only the systematic deviations. Still, these linear measures
tained for each traffic measure are compared with a case-control are strongly influenced by infrequently occurring high concentra-
(CC) that did not include the introduction of traffic measures tion values. Similar datasets have NMSE and FB equal zero. Finally,
with significant impact on the pollutant under study. MAXr (eq. (6)) is the ratio between the maximum values of two
In the first step, the main meteorological predictors of the datasets. Similar datasets have a MAXr equal 1.0.
pollutant concentrations (C) under study are identified. For
N  
example, for PM2.5 concentrations the ambient temperature (T) and 1 X
relative humidity (RH) can be the main predictors (Tai et al., 2010).
MBE ¼ Vðy
1Þi
Vyi [1]
N i¼1
Only the days with the minimum level of data captured (efficiency
higher than 75%) for all the key variables defined are analysed (e.g. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u N  2
for PM2.5 will be T, RH and C). This criterion must be used in order to u1 X
obtain a representative pattern of the data along the time. RMSE ¼ t Vyi
Vðy
1Þi [2]
N i¼1
Next, days with similar environmental conditions are identified.
Such days are identified by controlling the two main confounding
2 3
factors affecting the air quality management: meteorology and 2
emissions. Therefore, first, the variability of meteorological condi- 6 NðRMSEÞ 7
IOA ¼ 1
4   2 5 [3]
tions for the key meteorological predictors previously identified in PN    
i¼1 Vyi
Vy
1  þ Vðy
1Þi
Vy
1 
the first step is checked. For each hour (i) of each day (N), the values
observed (Vyi ) for the reference year (y) are compared with the  
values observed one year before (Vðy
1Þi ) (Fig. 1). Such comparison Vy
1
Vy
is accomplished by analysing several statistical functions that allow FB ¼   [4]
measuring the degree of similarity among the days. In this work, we 0:5* Vy
1 þ Vy
use the average (V), the standard deviation (SD), the average trend
of error (MBE), the root mean square error (RMSE), the coefficient of  2
correlation (r) and the index of agreement (IOA). These statistical Vy
1
Vy
NMSE ¼ [5]
functions were selected since they are frequently used and rec- Vy
1 *Vy
ommended in studies of environmental impact assessment to
evaluate the variability of meteorological conditions (e.g. Hurley,
MAXy
2008; Tang et al., 2009). Finally, to guarantee there were similar MAXr ¼ [6]
levels in the background concentrations in the previous day, the MAXy
1
daily average concentration ratio is analysed by comparing the
daily average concentration of that day (Vy ) with the average daily
value recorded one year before (Vy
1 ).
The last step includes the impact assessment of the traffic 2.2. Air quality traffic-related measures
measure on air pollutant concentrations for the days with similar
environmental conditions (‘similar days’). In this case, besides V, SD Two air quality traffic-related measures implemented in the city
and the IOA, the normalised mean square error (NMSE), the frac- of Beijing are selected to assess the impact on PM2.5 concentrations
tional bias (FB) and the ratio between maximums (MAXr ) are also using the methodology proposed in section 2.1. The first one (M1)
used to perform such analysis. The impact on high concentrations is has restricted the circulation of yellow-labelled vehicles, while the
assessed by computing the number of exceedances to the daily second one (M2) has restricted the circulation of heavy duty ve-
limit of human health protection. These functions are selected since hicles. Both measures are selected due to the high impact on the
they are frequently used and recommended in studies of environ- levels of PM2.5 emissions of road traffic: green-label measures emit
mental impact assessment to analyse the variability of air con- 25e90% less PM2.5 than light-duty yellow-labelled vehicles and
centrations in the atmosphere (e.g. Hurley, 2006; Mishra et al., 80e99% less PM2.5 than heavy-duty yellow-labelled vehicles (Shao
2015; Borrego et al., 2016). et al., 2014). Previous studies also demonstrate the impact of similar
V gives the average ‘central’ value of a dataset, while SD quan- measures on this pollutant (Boogaard et al., 2012; Holman et al.,
tified the dispersion of that set around this 'central' value. MBE (eq. 2015).
(1)) describes the deviation trend and the value obtained is related Measure 1 (M1), introduced on 1st October 2009, restricted the
with the magnitude of the analysed variable. An MBE negative yellow-labelled vehicles to drive within the Sixth-Ring Road of
value is achieved when Vy-1 is lower than Vy. On the other hand, Beijing (BENDIBAO, 2013). This measure was an enlargement of a
RMSE (eq. (2)) measures the magnitude differences between two measure introduced in 2003 that had restricted yellow-labelled
datasets. A value equals zero indicates equal datasets. According to vehicles to drive inside the Second-Ring Road of the city. From
Willmott (1982), this is one of the best measures to establish the 11th October 2014, yellow-labelled vehicles registered outside of
difference between two datasets. The IOA (eq. (3)) allows the Beijing were also forbidden to run within Beijing's administrative
analysis of the changes of one dataset against the other. Similar area (BJLZJ, 2014).
822 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828
Fig. 1. Methodological scheme of the temporal comparison of an air quality traffic-related measure.
Measure 2 (M2) was introduced on 11th April 2014. With this
measure, local freight vehicles (heavy and light) were restricted to
drive within Fifth-Ring Road of Beijing, an extension of the daylight
(between 6 a.m. and 11 p.m.) Fourth-Ring Road restriction imple-
mented in 2004 (BJLZJ, 2014). Furthermore, freight vehicles regis-
tered outside of Beijing were banned to drive within the Sixth-Ring
Road from 6 a.m. to 12 a.m., an extension of the Fifth-Ring road
restriction introduced on 3rd November 2009 (BJJTGL, 2009; BJLZJ,
2014).
Fig. 2 presents a spatial-temporal illustration of the imple-
mentation of these two air quality traffic-related measures in
Beijing.
2.3. Air quality and meteorological data
Hourly PM2.5 concentrations (mg.m
3) were collected by an
automatic urban monitoring air quality station. This is a back-
ground station, which typically represents an area with a radius
ranging between 100 m and 1 km (Larssen et al., 2012). The station
is located in the Chaoyang District, approximately 8 km east of the
city centre of Beijing city (Tiananmen Square) (Area 1 in Fig. 3).
Prior to 16th February 2009, 11 p.m. BJT, the station was located at
No. 3 Xiu Shui Bei Jie (39.31, 116.44 ) (A1i of Fig. 3), between the
Second-Ring Road and the Third-Ring Road; after that day, it was
moved to No. 55 An Jia Lou Road (39.95 , 116.47 ) (A1ii in Fig. 3),
between the Third-Ring Road and the Fourth-Ring Road. These
places are at 5 km from each other. A mix of a residential, com-
mercial and office buildings (with 2e20 floors) with green parks
characterises both these areas. The station is managed by the
United States Environmental Protection Agency (USEPA).
Hourly values of T, RH, WS (Wind Speed) and P (Pressure) were
collected in Beijing as well. For this purpose, two meteorological
stations were used: station 1 (S1), an urban station located in the
United States American embassy managed by the USEPA (Area 1 of
Fig. 3); and station 2 (S2), a mesoscale station located in the In-
ternational Airport of Beijing managed by the China Meteorological
Administration (Area 2 of Fig. 3). The stations were located at 8 km
and 25 km east of the city centre of Beijing city (Tiananmen Square),
respectively. According to WMO (2014), an urban meteorological
station is representative of scales ranging from one to several kil-
ometres, while a mesoscale meteorological station could represent
tens of kilometres in extent.
All data were collected between 2008 and 2015. This period
corresponds with the implementation of measures M1 and M2.
Fig. 3 shows the study area with the location of the main city traffic
lines and areas used for data collection, while Table 1 summarises
the information collected in this study area.
3. Results and discussion
To assess the methodology proposed, the impact of two distinct
air quality traffic-related measures implemented in the city of
Beijing on PM2.5 concentrations is investigated. First, to mitigate the
influence of environmental conditions, a set of similar days is
identified based on the pattern analysis of the levels of PM2.5
concentration and meteorological conditions. For those days, the
impact on PM2.5 concentrations is checked by analysing several
statistical functions. Since no statistical differences (p < 0.001) were
found between the overlapping meteorological data collected in
the meteorological stations S1 and S2, for the impact assessment of
M1, the meteorological data from S1 is used, while for M2, the
meteorological data from S2 is applied.
3.1. Identification of the ‘similar days’
A pre-analysis of the efficiency (<75%) on the data collection of
the key variables (C, T and RH) excludes 23.9% of the days analysed
between 2008 and 2015. During the period of M1 and M2 evalua-
tion, 173 (47.4%) and 49 (13.4%) of the days are excluded, respec-
tively. Note that, due to the inoperability of the air quality station
between 6th November of 2008 and 17th February of 2009, the
period prior to the implementation of M1, the number of excluded
days is very high, which limits the temporal coverage of the study
for this measure.
After this step, the datasets (Vy-1 and Vy) of each measure (M1
and M2) are compared. Since previous studies demonstrate that T
and RH are the key meteorological predictors of PM2.5 concentra-
tions (Tai et al., 2010), the comparison is done for these two vari-
ables. Besides the analysis of some statistical functions (V, SD, MBE
and RMSE) used to identify possible outliers in the daily variation of
the key meteorological variables (T and RH), the r and IOA statistics
are applied with the following cut-off limits: r0.75 and IOA0.5.
In addition, the level of background concentrations observed for
the previous day is also checked. To guarantee similar concentra-
tions levels, the following cut-off limit is used: 0.75Vy =Vy
1 1.25.
Descriptive statistics selected to assess the variability of the
meteorological conditions (r, IOA, MBE and RMSE) and background
concentrations (Vy =Vy
1 ) obtained for each measure (M1 and M2)
are summarised in Table 2 and Figure S2. The periods considered to
study the impact of each measure, M1 and M2, are highlighted at
light and dark grey, respectively.
The meteorological analysis show a similar pattern between the
 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828 823

Fig. 2. Spatial-temporal vision of the air quality traffic-related measures under study implemented in the city of Beijing (2008e2015).

Fig. 3. Location of the main arterials and points used for data collection in Beijing city.

period previously defined to evaluate each measure (M1 and M2)
and the remaining period of the study (2008e2015). In this anal-
ysis, a high percentage of the days used to study the measures’
impact shows low correlation coefficient values (r < 0.75) for T and/
or RH (M1: 109 days, M2: 165 days). Around 50% of these days are
rejected when applying the aforementioned cut-off limits defined
simultaneously for both meteorological variables (M1: 54 days, M2:
62 days). An individual analysis of such meteorological variables (T
and RH) suggests that data are more affected by the variability due
to RH (M1: 103 days, M2: 155 days) than T (M1: 60 days, M2: 72
days). Besides the rejected days due to the evaluation of r, the IOA
analysis leads to removing two more days for M1 and six more days
for M2. In both cases, the rejection of the additional days occurs due
to the high variability of T. Nevertheless, this last criteria (IOA<0.5)
824 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828

Table 1
Air quality and meteorological variables used in the study.

Variable Abbreviation Frequency Period Units Samples(N)

Air quality PM2.5 concentration CPM2.5 Hour 01/01/2008-30/11/2015 mg.m
3 61,294

PM2.5 exceedances EPM2.5 Daily n.º exceedances 2453

Meteorology Station 1 Temperature TS1 Hour 01/01/2008-10/08/2014 C 56,272
Relative Humidity RHS1 % 56,282
Wind speed WSS1 m.s
1 56,273
Pressure PS1 mbar 56,272

Station 2 Temperature TS2 Hour 01/01/2014-30/11/2015 C 16,775
Relative Humidity RHS2 % 16,775
Wind speed WSS2 m.s
1 16,775
Pressure PS2 mbar 16,775

has a lower impact in the identification of the days than the cor- assess the impact of M1 and 19 days used to assess the impact of
relation factor (M1: f(r): 109 days, f(IOA): 11 days; M2: f(r): 165 M2. These days are equally distributed throughout all seasons,
days, f(IOA): 32 days), which suggests a low variability of the enabling the study on the impact of each measure for the whole
average values of Vy-1 around the average values of Vy. year.
For T and RH, similar average and standard deviation values of r,
IOA, MBE and RMSE between M1 and M2 are found. Such values are 3.2. Evaluation of the measures impact
within the acceptable values usually obtained for these meteoro-
logical variables (Fontes, 2010). A detailed analysis shows a good Fig. 4 presents a set of descriptive statistics (Vn =Vn
1 , SDn =SDn
1 ,
agreement for all statistical functions, particularly for r and IOA IOA, NMSE, FB and MAXr) that allow to assess the daily variability of
(Table 2 and Figure S2). For these two functions, average values PM2.5 concentrations with the implementation of M1 and M2, and
higher than 0.9 are found for both traffic measures and periods. Table 2 shows the average values obtained for each of these sta-
Similar RMSE are also found for both measures: T ranges from 3.2 to tistics. To compare these trends, a time-period where the new
5.9, while RH ranges from 0.2 to 0.3. The MBE ranges from 1.8 measures were not yet applied, from 1st January of 2011 to 31st
to
1.1 for T and 0.3 to 0.2 for RH. December to 2011, is selected as case-control (CC). Only the ‘similar
Besides the variability of T and RH, the variability of P and WS days’ previously identified are used in this analysis. To better
was also checked. For that purpose, V and SD are analysed. recognise trends, moving averages are applied (Fig. 4).
Figure S3, in the Supplements section, displays the variability of V Distinct patterns of the descriptive statistics are found for M1
for the multiple meteorological variables (T, RH, P and WS), and M2. Although at first glance the IOA average values obtained
comparing the values observed during a reference period (Vy) with during the implementation of both measures suggest little varia-
the values observed one year previously (Vy-1). Similar average tions of the average PM2.5 concentrations through time (VIOA ¼ 0.8
values are found for the P along the eight analysed years. WS for both measures), an analysis of the dispersion values of all de-
variation shows some differences during the implementation of M1 viations highlights some differences. The variations of NMSE of each
(middle April until September). Still, through this period, the values measure with respect to the case-control are 0.0 and 0.6, for M1 and
observed after the implementation of this measure are consistently M2 respectively (Table 2). The daily variabilities also highlight
higher than the values observed before it. differences between the impact on PM2.5 concentrations of these
Finally, to control the influence of background concentrations, two measures (Fig. 4). M1 shows no significant deviations in the
the levels of PM2.5 in the previous day to the day under analysis are trends of these statistics when compared with the case-control.
analysed. In addition to the days excluded due to meteorology in- This first analysis suggests that M1 has an almost null impact on
fluence, 64 more days for M1 and 126 days for M2 are excluded PM2.5 concentrations. Furthermore, their capacity to improve the
after applying this criterion. As a result, the overall analysis allows air quality during hard dispersion conditions is not demonstrated.
identifying 145 days with similar environmental conditions For M2, the case is different. The variabilities of some statistical
distributed between 2008 and 2015. From these, 16 days are used to trends (Vn =Vn
1 , MAXr and NMSE) highlight significant variations
during the winter months. Even though, between 2008 and 2015,
high levels of PM2.5 concentrations are consistently recorded dur-
Table 2 ing the cold months, particularly from November to February, with
Summary of descriptive statistics for the background concentrations and the key
meteorological variables (T and RH) for the ‘similar days’ selected to assess the im-
the implementation of M2 such variation is even higher. This sug-
pacts of M1 and M2. gests an apparent ineffectiveness of this measure to reduce the high
levels of PM2.5 concentrations observed, especially during the
M1 M2 Control period All
winter season. This indicates that M1 has a low impact on PM2.5
CPM2.5 V y =V y
1 1.2 ± 0.7 1.2 ± 1.2 1.0 ± 0.6 1.1 ± 0.8 concentrations, while for M2 the benefits are not very clear due to
SDy/SDy-1 1.2 ± 0.7 1.1 ± 0.6 1.0 ± 0.7 1.1 ± 0.7 the increase of the maximum concentration, which can contribute
IOA 0.8 ± 0.1 0.8 ± 0.2 0.9 ± 0.1 0.8 ± 0.2
to quickly raise the degradation of the air quality levels in Beijing.
NMSE 0.8 ± 0.1 1.4 ± 1.3 0.8 ± 1.0 1.1 ± 1.9
FB
0.1 ± 0.5 0.2 ± 0.8 0.1 ± 0.6 0.1 ± 0.7 The particularities of each measure can explain the found dif-
MAXr 1.1 ± 0.7 1.0 ± 0.6 1.0 ± 0.6 1.0 ± 0.6 ferences. With the implementation of M1, drivers were forced to
T r 0.9 ± 0.0 0.9 ± 0.0 0.9 ± 0.0 0.9 ± 0.0 replace the high polluting vehicles (yellow-labelled vehicles) by
IOA 0.9 ± 0.1 1.0 ± 0.0 1.0 ± 0.1 1.0 ± 0.1
new ones. Therefore, if, as expected, the mobility patterns of drivers
MBE 1.8 ± 3.1
1.1 ± 2.8
0.9 ± 2.7
0.2 ± 3.0
RMSE 5.9 ± 3.6 3.2 ± 1.6 5.6 ± 3.3 4.0 ± 2.5
were not affected by this measure, a homogenous cut-off of the
RH r 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 0.9 ± 0.1 emissions is observed. As a result, at air quality level, a decrease of
IOA 0.9 ± 0.1 0.9 ± 0.0 1.0 ± 0.0 1.0 ± 0.0 the concentration levels of primary pollutants, as PM2.5, is ex-
MBE 0.1 ± 0.2 0.0 ± 0.2 0.0 ± 0.1 0.0 ± 0.2 pected. Concerning the implementation of M2, the case is different.
RMSE 0.3 ± 0.1 0.2 ± 0.1 0.2 ± 0.1 0.2 ± 0.1
With this measure, a change of the PM2.5 concentration pattern in
 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828 825

Fig. 4. Descriptive statistics (Vn =Vn
1 , SDn =SDn
1 , IOA, NMSE, FB and MAXr) of PM2.5 concentrations (mg.m
3) between 2009 and 2015. In this analysis, a moving average of 6 days is
applied using only the ‘similar days’ identified.
826 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828
the atmosphere is expected. This happens due to two main reasons:
(i) drivers of duty vehicles were forced to drive only during the
night; and (ii) duty vehicles are the main source of PM2.5 road traffic
emissions. Therefore, with the implementation of M2, an important
fraction of road traffic PM2.5 emissions is shifted to the night-time.
At first glance, we can expect a decrease of the high concentration
values during the day and an increase during the night. Never-
theless, during night-time the capacity of atmosphere dispersion
declines, given that with the drop of T the height of the boundary
layer decreases (Seinfeld and Pandis, 2016). This promotes the in-
crease of the concentration levels of pollutants in the atmosphere.
Thus, even though during the night-time the concentration value is
lower than during the day and the number of population exposed is
limited, the impact of this phenomenon during the early morning
and late afternoon (when people are starting and ending their
commuting movements, respectively) can be critical. The reduction
of the atmospheric capacity for transport and dispersion of the
pollutant emissions during night-time can also contribute to
increasing the daily average values.
To quantify the daily effects of the traffic measures on the high
levels of PM2.5 concentrations, the interim target III (75 mg.m
3),
defined by the WHO (2005) for human health protection, and
adopted by China, is used. Two distinct analyses are conducted
using this limit. First, the difference between the percentage of days
in a month exceeding this limit is computed. Then, the monthly
average of the PM2.5 concentrations (mg.m
3) is estimated. For both
cases, data are computed for the period between 2008 and 2015
using the ‘similar days’ previously identified. Values observed
during the reference period (Vy) are compared with the values
obtained one year previously (Vy-1). Fig. 5 shows the obtained
results.
For the case-control, the average PM2.5 concentrations increased
by 18% and the number of exceedances of PM2.5 to the interim
target III (75 mg.m
3) increased by 27% (from 98.6 mg.m
3 to
119.8 mg.m
3 and 54%e75%, respectively). On the other hand, dur-
ing the period of evaluation of each measure, a different trend is
observed. For M1, the average concentrations decreased by 5%, from
95.4 mg.m
3 to 91.1 mg.m
3, while for M2, the levels increased by
16%, from 102.4 mg.m
3 to 121.6 mg.m
3. The impact on the average
number of exceedances is similar for both measures, with a
decrease of 6% and 8% of the daily exceedances for M1 and M2,
respectively. These results suggest that even though the values
have increased for M2, its comparison with the case-control dem-
onstrates that, although with little effect, it has the capacity to
mitigate the impact of this pollutant. These results are in accor-
dance with the previous results that highlight the little effect of
traffic measures on the levels of PM2.5 concentrations of Beijing
(Sun et al., 2014). Still, the high variability of the values along the
period of study (Fig. 5), influenced by the low number of ‘similar
days’ in each month, did not allow to identify any clear seasonal
trend. The dynamics changes of the city may have contributed to
this variability. Moreover, the adaptation of drivers to other travel
modes, the continuous growth of the city and/or the introduction of
new measures can also be pointed out as some of the reasons that
contribute to eliminating the benefits of the introduction of some
measures.
In spite of the results of the PM2.5 concentrations being collected
from an air quality station located inside the area affected by the
restriction measures studied, scant evidence of their impact on
PM2.5 concentrations can be observed. This can occur for several
reasons. First, we are working with an urban background air quality
station. This station type is more representative of the air quality
levels of a city, but the impact from a specific emission source, as
the road vehicles, can be hard to identify. Second, the low impact of
air quality traffic-related measures can be related to the low
contribution of the road traffic sector on PM2.5 levels in Beijing.
Previous source apportionment studies conducted in the city sug-
gest that road traffic emissions only contribute between 10 and 20%
to the overall road traffic PM2.5 emissions (Huang et al., 2014; Li
et al., 2015; Yang et al., 2016), while the contribution of second-
ary particles ranges from 30 to 60% (Huang et al., 2014). Moreover,
Yang et al. (2015) showed in a review study that, between 2008 and
2015, the total vehicle PM2.5 emissions in Beijing was almost con-
stant, which suggests that the air quality traffic-related measures
implemented during these years only slowed down the impact
produced by the general increase of traffic flow in the city (BJTRC,
2017).
4. Conclusion
In this study, a statistical methodology facilitating the under-
standing of the real impact of air quality traffic-related measures on
air quality is proposed. The method compares the air quality levels
observed after the implementation of a traffic restriction measure
with the values observed one year before. Short and long-term
impact are evaluated considering not only the average variability
along time, but also the impact on different levels of concentration.
To control the influence of the main confounding factors, such
analysis is only done for days with similar environmental condi-
tions (‘similar days’). Those days are selected by analysing the
patterns of some meteorological variables that are specified ac-
cording to the pollutant under study. The similarity of background
conditions is also checked for the previous day. The variability of
the pollutant concentrations is analysed for the ‘similar days’
identified using several specific statistical functions.
To assess the methodology proposed, the impact on PM2.5
concentrations of two traffic restriction measures implemented in
the city of Beijing are evaluated: M1 was implemented in 2009,
restricting the circulation of yellow-labelled vehicles, while M2 was
implemented in 2014, restricting the circulation of heavy-duty
vehicles. A time-period of one year when the new measures had
not yet been applied is used as case-control.
A pre-analysis of the data collected in Beijing allows identifying
a total of 145 days with similar environmental conditions between
2008 and 2015, equally distributed along the months with available
data. From these, 16 days are used to assess the impact of M1 and 19
days are used to assess the impact of M2. The pairs of days selected
have analogous atmospheric dispersion conditions (r  0.75 and
IOA0.5) for the key meteorological variables that mostly influence
the variations of PM2.5 concentrations in the atmosphere: T and RH
(Tai et al., 2010), as well as similar background concentrations in the
previous day to day under evaluation (0.75Vy =Vy
1 1.25). Thus,
we consider such days comparable. Note, however, that we cannot
remove all the confounding factors that may mislead the impact
assessment of the measures. Since emissions and meteorology act
as the main confounding factors, we conclude that by applying this
methodology, we can better understand the impact of a traffic
measure on air quality by identifying some signs in the trends of air
pollution often hidden within environmental variability.
An analysis of the ‘similar days’ reveals distinct impacts on PM2.5
concentrations for M1 and M2. The average monthly concentra-
tions' and exceedances' trends show that both traffic measures
have the capacity to slow-down the increase of the levels of PM2.5
concentrations, even though the analysis of the daily variability is
not clear. The variability of descriptive statistics suggests that M1
has a low impact on the levels of average PM2.5 concentrations, and
its capacity to improve the concentrations levels of this pollutant
during the worst season is not fully demonstrated. For M2, the case
is different. Although the number of exceedances had decreased,
the maximum values increased during the worst season. Two
 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828
Fig. 5. Percentage of days exceeding the interim target III (75 mg.m
3) for PM2.5 concentrations, and average PM2.5 concentrations (mg.m
3). Monthly average differences are
computed comparing the values of a reference year (Vy), with the values of the prior year (Vy-1). Only ‘similar days’ observed between 2009 and 2015 are used.
factors play important roles for this result: (i) the transference of
emissions to night-time; and (ii) the low capacity of the atmo-
sphere for pollutant transport and dispersion during colder periods
(night and winter). This suggests that M1 has a low but positive
impact on PM2.5 concentrations, and its benefits may have been
masked due to the influence of negative external factors induced by
the continuous population growth of the city. Such trends are ex-
pected to be clearer if black carbon concentration (not available for
this study) is used to analyse this type of measures (Holman et al.,
2015). For M2, the results suggest that this measure has contributed
to mitigating some of the impacts on PM2.5 concentrations; still, the
maximum levels of concentration may have been negatively
affected. This can contribute to degrading severely the air quality
during particular days, which reinforces the need for the traffic
restriction measures to be better designed and assessed. Although
measures to reduce emission of traffic related particles can be
relevant, additional measures are needed to abate PM in the city of
Beijing. The urban policies need to be focused on emission sources
with higher contribution to the overall emissions and the impact of
such measures must be assessed using robust models that control
the influence of meteorology.
Acknowledgments
This work was financed by the Erasmus Mundus Program within
the project LOTUSþ and by Portuguese National Funds through FCT
e Fundaça ~o para a Cie
^ncia e Tecnologia (Portuguese Foundation for
Science and Technology) as part of the Project UID/Multi/04546/
2013. This research is also partly funded by the National Natural
Science Foundation of China (Project No. 41571147). The authors
also acknowledge the U.S. Department of State for the data pro-
vided to conduct the study (Ref. LM15PD0252). Ta ^nia Fontes would
like to personally thank FCT for the Pos-Doc grant SFRH/BPD/
109426/2015.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.envpol.2018.04.061.
References
Batterman, S., Xu, L., Chen, F., Chen, F., Zhong, X., 2016. Characteristics of PM2.5
concentrations across Beijing during 2013e2015. Atmos. Environ. 145, 104e114.
Barros, N., Fontes, T., Manso, C., 2013. How wide should be the adjacent area to an
827
urban motorway to prevent potential health impacts from traffic emissions?
Transport. Res. Pol. Pract. 50, 113e128.
BENDIBAO, 2013. Yellow cars: what does it mean? [in Chinese] Retrieved from
BENDIBAO. http://bj.bendibao.com/zffw/201394/115874.shtm.
BJJTGL, 2009. Beijing public security: bureau notices. Retrieved from Beijing public
security bureau on adjusting the other towns Beijing: notices truck traveling
road range [in Chinese]. http://bjjtgl.gov.cn/jgj/bjsgajtg/335154/index.html.
BJLZJ, 2014. Beijing announced to take traffic management measures on the part of
the vehicle to reduce emissions notices [in Chinese]. Retrieved from. http://glcx.
bjlzj.gov.cn/bjglwww/cxts/03/16658.shtml.
BJGOV, 2014. Transport management measure in order to mitigate pollutant
emission [in Chinese]. Retrieved from. http://glcx.bjlzj.gov.cn/bjglwww/cxts/
03/16658.shtml.
BJTRC, 2017. Beijing Transport Annual Report (2003-2016) [in Chinese]. Retrieved
from. http://www.bjtrc.org.cn/JGJS.aspx?id¼5.2&Menu¼GZCG.
BJSTATS, 2016. Beijing statistical information net [in Chinese]. Retrieved 3 2016,
from. http://www.bjstats.gov.cn/tjsj/tjgb/ndgb/201603/t20160329_346055.
html.
Boogaard, H., Janssen, N.A.H., Fischer, P.H., Kos, G.P.A., Weijers, E.P., Cassee, F.R., van
der Zee, S.C., Hartog, J.J., Meliefste, K., Wang, M., Brunekreef, B., Hoek, G., 2012.
Impact of low emission zones and local traffic policies on ambient air pollution
concentrations. Sci. Total Environ. 435e436 (2012), 132e140.
Borrego, C., Amorim, J., Tchepel, O., Dias, D., Rafael, S., Sa, E., Pimentel, C., Fontes, T.,
Fernandes, P., Pereira, S.R., Bandeira, J.M., Coelho, M., 2016. Urban scale air
quality modelling using detailed traffic emissions estimates. Atmos. Environ.
131, 341e351.
Cai, W., Li, K., Liao, H., Wang, H., Wu, L., 2017. Weather conditions conducive to
Beijing severe haze more frequent under climate change. Nat. Clim. Change 7,
257e262.
Chen, R., Zhao, Z., Kan, H., 2013. Heavy smog and hospital visits in Beijing, China.
Am. J. Respir. Crit. Care Med. 188, 1170e1171.
Chen, Y., 2015. Commuting time survey in 50 cities: average 19.2 km one way in
Beijing. Chinanews.com. http://www.chinanews.com/sh/2015/01-26/7005909.
shtml.
Chen, Y., Schleicher, N., Chen, Y., Chai, F., Norra, S., 2014. The influence of govern-
mental mitigation measures on contamination characteristics of PM2.5 in
Beijing. Sci. Total Environ. 490, 647e658.
Chu, M., Sun, C., Chen, W., Jin, G., Gong, J., Zhu, M., Yuan, J., Dai, J., Wang, M., Pan, Y.,
Song, Y., Ding, X., Guo, X., Du, M., Xia, Y., Kan, H., Zhang, Z., Hu, Z., Wu, T.,
Shen, H., 2015. Personal exposure to PM2.5, genetic variants and DNA damage: a
multi-center population-based study in Chinese. Toxicol. Lett. 235, 172e178.
Cyrys, J., Peters, A., Soentgen, J., Wichmann, H.-E., 2014. Low emission zones reduce
PM10 mass concentrations and diesel soot in German cities. J. Air Waste Manag.
Assoc. 64 (4), 481e487.
DieselNet, 2018. China: Cars and Light Trucks in Emission Standards. https://www.
dieselnet.com/standards/cn/ld.php.
Dong, H., Zeng, L., Hu, M., Wu, Y., Zhang, Y., Slanina, J., Zheng, M., Wang, Z.-F.,
Jansen, R., 2012. Technical note: the application of an improved gas and aerosol
collector for ambient air pollutants in China. Atmospheric Chemical Physics 12,
10519e10533.
Ding, Y.H., Liu, Y.J., 2014. Analysis of long-term variations of fog and haze in China in
recent 50 years and their relations with atmospheric humidity. Sci. China Earth
Sci. 57, 36e46.
Fontes, T., 2010. Impacte da qualidade do ar urbana na saúde pública: O caso da
cidade do Porto. University of Aveiro, Aveiro.
Fontes, T., Li, P., Barros, N., Zhao, P., 2017. Trends of PM2.5 concentrations in China: a
long term approach. J Environ Manage. J 196, 719e732.
828 T. Fontes et al. / Environmental Pollution 239 (2018) 818e828
Guo, H., Wang, Y., Zhang, H., 2017. Characterization of criteria air pollutants in
Beijing during 2014e2015. Environ. Res. 154, 334e344.
Hao, J., Hu, J., Fu, L., 2006. Controlling vehicular emissions in Beijing during the last
decades. Transport. Res. Pol. Pract. 639e651.
Harrison, R.M., Beddows, D.C., 2017. Efficacy of recent emissions controls on road
vehicles in Europe and implications for public health. Scientic Reports 7 (1152).
https://doi.org/10.1038/s41598-017-01135-2.
Holman, C., Harrison, R., Querol, X., 2015. Review of the efficacy of low emission
zones to improve urban air quality in European cities. Atmos. Environ. 111,
161e169.
Huang, J., Deng, F., Wu, S., Guo, X., 2012. Comparisons of personal exposure to
PM2.5 and CO by different commuting modes in Beijing, China. Sci. Total En-
viron. 425, 52e59.
Huang, R., Zhang, Y., Bozzetti, C., Ho, K.-F., Cao, J.-L., Han, Y., Daellenbach, K.R.,
Slowik, J.G., Platt, S.M., Canonaco, F., Zotter, P., Wolf, R., Pieber, S.M., Bruns, E.A.,
Crippa, M., Ciarelli, G., Piazzalunga, A., Schwikowski, M., Abbaszade, G.,
Schnelle-Kreis, J., Zimmermann, R., An, Z., Szidat, S., Baltensperger, U.,
Haddad, I., Pre vo^t, A.S.H., 2014. High secondary aerosol contribution to partic-
ulate pollution during haze events in China. Nature 514, 218e222.
Huo, H., Yao, Z., Zhang, Y., Shen, X., Zhang, Q., He, K., 2012. On-board measurements
of emissions from diesel trucks in five cities in China. Atmos. Environ. 54,
159e167.
Hurley, P., 2006. An evaluation and inter-comparation of AUSPLUME, AERMOD and
TAPM for seven field datasets of points source dispersion. Clean Air Environ.
Qual. 40 (1), 45e50.
Hurley, P., 2008. Development and Verification of TAPM. In: Air Pollution Modelling
and its Application XIX. Springer Netherlands, pp. 208e216.
Kimbrough, S., Baldauf, R., Hagler, G., Shores, R., Mitchell, W., Whitaker, D.,
Croghan, C.W., Vallero, D., 2013. Long-term continuous measurement of near-
road air pollution in Las Vegas: seasonal variability in traffic emissions
impact on local air quality. Air Quality Atmosphere and Health 6, 295e305.
Larssen, S., Sluyter, R., Helmis, C., 2012. Criteria For Euroairnet - The Eea Air Quality
Monitoring and Information Network. Technical Report No. 12. European
Environmental Agency, Copenhagem.
Lai, D., 2015. China Labor Market Development Report [in Chinese]. Beijing Normal
University, Beijing.
Li, P., Xin, J., Wang, Y., Wang, S., Li, G., Pan, X., Liu, Z., Wang, L., 2013. The acute effects
of fine particles on respiratory mortality and morbidity in Beijing, 2004e2009.
Environ. Sci. Pollut. Res. Int. 20 (9), 6433e6444.
Li, X., Zhang, Q.,Y.,Z., Zheng, B., Wang, K., Chen, Y., Wallington, T.J., Han, W., Shen, W.,
Zhang, X., He, K., 2015. Source contributions of urban PM2.5 in the Beijing-
Tianjin-Hebei region: changes between 2006 and 2013 and relative impacts
of emissions and meteorology. Atmos. Environ. 123, 229e239.
Maricp, M., 2013. Monitoring motor vehicle PM emissions: an evaluation of three
portable low-cost aerossol instruments. Aerosol Sci. instrum. 47 (5), 564e573.
Minghao, W., 2011. Beijing provides both government subsidy and company coupon
to promote old vehicle eliminations [in Chinese]. People's daily. http://politics.
people.com.cn/GB/14562/15289712.html.
Mishra, D., Goyal, P., Upadhyay, A., 2015. Artificial intelligence based approach to
forecast PM2.5 during haze episodes: a case study of Delhi, India. Atmos. En-
viron. 102, 239e248.
Okuda, T., Matsuura, S., Yamaguchi, D., Umemura, T., Hanada, E., Orihara, H.,
Tanaka, S., He, K., Mab, Y., Cheng, Y., Liang, L., 2011. The impact of the pollution
control measures for the 2008 Beijing Olympic Games on the chemical
composition of aerosols. Atmos. Environ. 45 (16), 2789e2794.
Ordo n ~ ez, C., Mathis, H., Furger, M., Henne, S., Hüglin, C., Staehelin, J., Pre vo
^t, A.S.H.,
2005. Changes of daily surface ozone maxima in Switzerland in all seasons from
1992 to 2002 and discussion of summer 2003. Atmos. Chem. Phys. 5,
1187e1203.
Raaschou-Nielsen, O., Andersen, Z., Beelen, R., Samoli, E., Stafoggia, M.,
Weinmayr, G., Hoffmann, B., Fischer, P., Nieuwenhuijsen, M.J., Brunekreef, B.,
Xun, W.W., Katsouyanni, K., Dimakopoulou, K., Sommar, J., Forsberg, B.,
Modig, L., Oudin, A., Oftedal, B., Schwarze, P.E., Nafstad, P., Faire, U.,
€
Pedersen, N.L., Ostenson, C.-G., Fratiglioni, L., Penell, J., Korek, M., Pershagen, G.,
Eriksen, K.T., Sørensen, M., Tjønneland, A., Ellermann, T., Eeftens, M.,
Peeters, P.H., Meliefste, K., Wang, M., Bueno-de-Mesquita, B., Key, T.J., Hoogh, K.,
Concin, H., Nagel, G., Vilier, A., Grioni, S., Krogh, V., Tsai, M.-Y., Ricceri, F.,
Sacerdote, C., Galassi, C., Migliore, E., Ranzi, A., Cesaroni, G., Badaloni, C.,
Forastiere, F., Tamayo, I., Amiano, P., Dorronsoro, M., Trichopoulou, A., Bamia, C.,
Vineis, P., Hoek, G., 2013. Air pollution and lung cancer incidence in 17 Euro-
pean cohorts: prospective analyses from the European Study of Cohorts for Air
Pollution Eff ects (ESCAPE). Lancet Oncol. 14 (813), 822.
Roy, A., Hu, W., Wei, F., Korn, I.C., Zhang, J., 2012. Ambient particulate matter and
lung functions growth in Chinese childrens. Epidemiology 22 (3), 464e472.
Seinfeld, J., Pandis, S., 2016. Atmospheric Chemistry and Physics: from Air Pollution
to Climate Change, third ed. EUA: Wiley.
Shao, Z., Wagner, D., Yang, Z., 2014. Costs and benefits of China 5/V and 6/vi in
guangdong province. ICCT report. June 16. Retrieved from. http://www.theicct.
org/costs-and-benefits-china-5v-and-6vi-standards-guangdong-provinceICCT
Report.
Shen, X., Yao, Z., Huo, H., He, K., Zhang, Y., Liu, H., Ye, Y., 2014. PM2.5 emissions from
light-duty gasoline vehicles in Beijing, China. Sci. Total Environ. 487, 521e527.
Sun, C., Zheng, S., Wang, R., 2014. Restricting driving for better traffic and clearer
skies: did it work in Beijing? Transport Pol. 32, 34e41.
Tai, A., Mickley, L., Jacob, D., 2010. Correlations between fine particulate matter
(PM2.5) and meteorological variables in the United States: implications for the
sensitivity of PM2.5 to climate change. Atmos. Environ. 44 (32), 3976e3984.
Tang, L., Miao, J.F., Chen, D., 2009. Performance of TAPM against MM5 at urban scale
during GOTE2001 campaign. Boreal Environ. Res. 14, 338e350.
USEPA, 1992. Protocol for Determining the Best Performing Model. USEPA, Wash-
ington D.C., USA.
Wang, H., Fu, L., Zhou, Y., Du, X., Ge, W., 2010a. Trends in vehicular emissions in
China's mega cities from 1995 to 2005. Environ. Pollut. 158, 394e400.
Wang, T., Xie, S., 2010. Assessment of traffic-related air pollution in the urban
streets before and during the 2008 Beijing Olympic Games traffic control
period. Atmos. Environ. 43, 5682e5690.
Wang, T., Nie, W., Gao, J., Xue, L.K., Gao, X.M., Wang, X.F., Qiu, J., Poon, C.N.,
Meinardi, S., Blake, D., Wang, S.L., Ding, A.J., Chai, F.H., Zhang, Q.Z., Wang, W.X.,
2010b. Air quality during the 2008 Beijing Olympics: secondary pollutants and
regional impact. Atmospheric Chemical Physics 10, 7603e7615.
Wang, Y., Yao, L., Wang, L., Liu, Z., Ji, D., Tang, G., Zhang, J., Sun, Y., Hu, B., Xin, J., 2013.
Mechanism for the formation of the January 2013 heavy haze pollution episode
over central and eastern China. Sci. China Earth Sci. 57 (1), 14e25.
Willmott, C.J., 1982. Some comments on the validation of model performance. Bull.
Am. Meteorol. Soc. 63, 1309e1313.
WHO, 2005. WHO Air Quality Guidelines for Particulate Matter, Ozone, Nitrogen
Dioxide and Sulfur Dioxide. World Health Organization, Geneva.
WMO, 2014. WMO Guide to Meteorological Instruments and Methods of Obser-
vation (the CIMO Guide), Part II e Observing Systems, Chapter 9 e Urban Ob-
servations. WMO-no. 8 (2014 edition). World Meteorological Organization,
p. 766.
Wu, Y., Wang, R., Zhou, Y., Lin, B., Fu, L., He, K., Hao, J., 2011. On-road vehicle
emission control in Beijing: past, present and future. Environ. Sci. Tecnol. 45 (1),
147e153.
Xu, H., Ley, Y., Haofang, C., 2011. Annual Report on Urban-rural Development of
Beijing [in Chinese]. Social Science Academic Press, Beijing.
Yanan, L., 2015. Metro usher in a new round of construction boom (policy inter-
pretation $ Focus) [in Chinenes]. People. Retrieved from. http://paper.people.
com.cn/rmrb/html/2015-07/03/nw.D110000renmrb_20150703_2-03.htm.
Yang, H., Chen, J., Wen, J., Tian, H., Liu, X., 2016. Composition and sources of PM2.5
around heating periods of 2013 and 2014 in Beijing: implications for efficient
mitigation measures. Atmos. Environ. 124 (B), 378e386.
Yang, Z., Wang, H., Shao, Z., Muncrief, R., 2015. Review of Beijing's Comprehensive
Motor Vehicle Emission Control Programs (White Paper). ICCT, Berlin.
Zhang, S., Wu, Y., Wu, X., Li, M., Ge, Y., Liang, B., Xu, Y., Zhou, Y., Liu, H., Fu, L., Hao, J.,
2014. Historic and future trends of vehicle emissions in Beijing, 1998-2020: a
policy assessment for the most stringent vehicle emission control program in
China. Atmos. Environ. 89, 216e229.
Zhao, P., 2015. Emissions of air pollutants from road transports in Beijing: facts,
factors and forecast. In: Planning for Sustainable Cities: Urban Challenges,
Policy Responses and Research Agenda (Pp. 145e154). New Zealand. National
Library of New Zealand.
Zhao, P., Zhang, X., Xu, X., Zhao, X., 2011. Long-term visibility trends and charac-
teristics in the region of Beijing, Tianjin, and Hebei, China. Atmostheric
Research 101 (3), 711e718.
Zhou, Y., Wu, Y., Yang, L., Fu, L., He, K., Wang, S., Hao, J., Chen, J., Li, C., 2010. The
impact of transportation control measures on emission reductions during the
2008 Olympic Games in Beijing, China. Atmos. Environ. 44, 285e293.
Zhuang, X., Wang, Y., He, H., Liu, J., Wang, X., Zhu, T., Ge, M., Zhou, J., Tang, G., Ma, J.,
2014. Haze insights and mitigation in China: an overview. J. Environ. Sci. 26 (1),
2e12.