Environmental Pollution 249 (2019) 1e12

Contents lists available at ScienceDirect

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

A tale of two cities: is air pollution improving in Paris and London?*

Anna Font a, b, *, Lionel Guiseppin c, Marta Blangiardo b, d, Ve

ronique Ghersi c,
a, b
Gary W. Fuller
a
School of Population Health & Environmental Sciences, Faculty of Life Sciences & Medicine, King's College London, UK
b
MRC PHE Centre for Environment and Health, UK
c
Observatoire de l'air en ^Ile-de-France, Airparif, France
d
Department of Epidemiology and Biostatistics, School of Medicine, Imperial College London, UK

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

Article history: Paris and London are Europe's two megacities and both experience poor air quality with systemic
Received 15 May 2018 breaches of the NO2 limit value. Policy initiatives have been taken to address this: some European-wide
Received in revised form (e.g. Euro emission standards); others local (e.g. Low Emission Zone, LEZ). Trends in NOX, NO2 and
10 January 2019
particulate matter (PM10, PM2.5) for 2005e2016 in background and roadside locations; and trends in
Accepted 11 January 2019
traffic increments were calculated in both cities to address their impact. Trends in traffic counts and the
Available online 14 January 2019
distribution in Euro standards for diesel vehicles were also evaluated. Linear-mixed effect models were
built to determine the main determinants of traffic concentrations. There was an overall increase in
Keywords:
Megacity
roadside NO2 in 2005e2009 in both cities followed by a decrease of ~5% year
1 from 2010. Downward
Trends in air pollution trends were associated with the introduction of Euro V heavy vehicles. Despite NO2 decreasing, at cur-
Diesel rent rates, roads will need 20 (Paris) and 193 years (London) to achieve the European Limit Value
Euro standards (40 mg m
3 annual mean). Euro 5 light diesel vehicles were associated with the decrease in roadside
Low emission zone PM10. An increase in motorcycles in London since 2010 contributed to the lack of significant trend in
PM2.5 roadside increment in 2010e16.
© 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction both pollutants in London (GLA , 2017). Diesel engines emit NOX in
the form of nitrogen monoxide (NO) and nitrogen dioxide (NO2).
Paris and London are Europe's two megacities with more than NO is rapidly oxidised to NO2 in the atmosphere through its reac-
10 million people living in each. Both cities experience poor air tion with ozone leading to large concentrations near to roads and
quality. They currently exceed the European Limit Value for nitro- ultimately in urban areas. Uncomplete combustion in diesel en-
gen dioxide (NO2) and particulate matter (PM) episodes are gines emit fine particles (particles with <2.5 mm). Particle emis-
frequent during winter and early spring (Bessagnet et al., 2005; sions are also emitted by non-exhaust sources such as
Favez et al., 2012; Rouïl et al., 2015; Macintyre et al., 2016; Petit resuspension, tyre-wear and brake-wear, which represent an
et al., 2017). Long-term exposure to poor air quality has been important fraction of coarse PM (particles with >2.5 mm) on roads
associated with premature deaths estimated to be ~55,000 annu- (Amato et al., 2016; Font and Fuller, 2016).
ally in France and ~50,000 in the United Kingdom (Guerreiro et al., Targeted policies to reduce vehicle emissions have been
2016). implemented at multiple scales. Some of these are European-wide
Road transport accounts for a large proportion of primary such as the introduction of the Euro emission standards on new
emissions of nitrogen oxides (NOX) and PM10 in both cities as vehicles. Euro norms were introduced in 1991 and new stages were
estimated by inventories: 73% (NOx) and 42% (PM10) in Paris; 61% defined over time with more stringent exhaust emission standards
(NOx) and 28% (PM10) in ^Ile-de-France (Airparif, 2016); and 50% for for new vehicles. However, real-world emissions in new cars did
not always align with expected type-approval tests. While test-
cycle NOX emissions decreased by 80% since 1992, the real-
driving emissions from diesel cars have increased ~20% (Carslaw
*
This paper has been recommended for acceptance by Eddy Y. Zeng. et al., 2016; Weiss et al., 2012). The after-treatment devices to
* Corresponding author. 150 Stamford Street, London, SE1 9NH, United Kingdom. reduce PM emissions in diesel vehicles were proved to increase
E-mail address: [email protected] (A. Font).

https://doi.org/10.1016/j.envpol.2019.01.040
0269-7491/© 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
2 A. Font et al. / Environmental Pollution 249 (2019) 1e12

primary NO2 emissions since the introduction of Euro 3/III diesel 2. Methods
vehicles, representing around 50% of the on-road emissions of NOX
(Carslaw et al., 2016; Grice et al., 2009; Weiss et al., 2012); and NO2 2.1. Air quality data
emissions in Euro 6a were still higher than those in Euro 2 (Carslaw
et al., 2016). Despite the effective reduction of on-road NOX emis- Hourly measurements of nitrogen oxides (NOX), nitrogen diox-
sions from Euro 6 diesel cars (Weiss et al., 2012), early Euro 6 diesel ide (NO2), particulate matter with aerodynamic diameter of less
cars showed large variability in urban performance and exceeded than 10 mm and 2.5 mm (PM10 and PM2.5) were obtained from Air-
the NOX standard limit (Degraeuwe and Weiss, 2017; O'Driscoll parif (Paris); and the London Air Quality Network (LAQN) and
et al., 2016; Weiss et al., 2012). The new Euro 6d-temp standard Defra's Automatic Urban and Rural Network (AURN) (London) from
for cars includes a real-world driving limit for NOX for the first time. 2005 to 2016. This data set comprised a total of 44 monitoring sites
In contrast to problems with diesels, on-road tests for gasoline Euro across the ^Ile-de-France region: 30 background locations and 14
3e5 light good vehicles (LGVs) were within emissions limits (Weiss roadside sites; and 130 monitoring sites in Greater London with 51
et al., 2011). background and 79 roadside sites. Note that not all locations
Other policies to reduce air pollutant concentrations include the measured all pollutants (Supplementary Table 1, Supplementary
introduction of Low Emission Zones (LEZs). The LEZ in London came Table 2). The distance to Paris or London's city centres was calcu-
into operation in February 2008 (for heavy good vehicles of more lated for each monitoring site setting the centre of Paris at the
than 12 tonnes in phase one) and July 2008 (for all other heavy good “Point ze
ro des routes de France” located in the parvis of Notre-
vehicles in phase two) setting a minimum Euro III PM emission Dame (48 51’12.25’’N, 2 20’55.63’’E); and at Charing Cross in
standard for these vehicles. Third and fourth phases were intro- London (5130’28.8’’N, 0 7’30’’W).
duced in January 2012 requiring larger vans, minibuses and other NOX (NO þ NO2) was measured in both cities by chemi-
specialist vehicles to meet Euro III standards for PM; and heavy luminiscence and regular calibrations enabled the traceability of
good vehicles standards were tightened to Euro IV. Euro III HGVs measurements to national metrological standards. In Paris, PM was
fitted with an approved filter or converted to an approved gas en- measured by TEOM (Tapered Element Oscillating Microbalance)
gine were allowed to continue to enter the LEZ. Buses operated by corrected by TEOM-FDMS (Tapered Element Oscillating Microbal-
Transport for London (TfL) had to meet Euro IV standard from ance - Filter Dynamics Measurement System) from 2005 to 2011.
December 2015. A LEZ based on Euro norm vehicle classification From 2012 onwards PM was measured by TEOM-FDMS which is
was introduced in Paris in September 2015 banned the circulation considered equivalent to the EU reference method based on 24-h
of Euro II heavy good vehicles, coaches and buses in inner Paris. A sampling and gravimetric analysis. PM10 in London was measured
second phase was introduced in July 2016 with the interdiction of by either TEOM-FDMS, by TEOM and or by MetOne BAM (Beta
pre-Euro 2 cars and motorcycles. A third phase introduced in July Attenuation Monitor). PM10 measurements by TEOM-FDMS were
2017 banning Euro 2 diesel cars and Euro III heavy good vehicles. All reference equivalent. PM10 concentrations by TEOM were con-
these restrictions are limited to weekdays from 8 a.m. to 8 p.m. verted to reference equivalent using the Volatile Correction Model
Other local measures include the introduction of emergency (VCM) (Green et al., 2009). PM10 measurements by BAM were
measures during pollution episodes in Paris (consisting in speed corrected to EU reference equivalent using a factor of 1/1.2 (DEFRA,
restrictions; banning heavy duty vehicles in inner Paris; free or 2010). PM2.5 measurements were measured by TEOM-FDMS in
reduced cost public transportation; free residential parking; alter- both cities. In Paris, the Airparif laboratory and instrumentation
nate number plate traffic circulation until 2016 and differentiated were certificated by the French accreditation committee (COFRAC)
circulation based on vehicle classification since 2017; and also re- for calibration and testing activities (European norm: ISO/CEI
strictions on wood burning and industrial emissions); and the 17025). In London, all instruments were subject to twice yearly
conversion to a pedestrian street of “Voie Georges-Pompidou” in audit tests by the National Physical Laboratory or Ricardo AEA
September 2016, that previously carried 43,000 vehicles every day. (London) which hold UKAS and ISO/CEI 17025 certifications.
Measures in London include the TfL retrofitting program of Euro III
buses with Selective Catalytic Reduction (SCR) in specific routes in
London; and the introduction of the Low Emission Bus Zone in 2017. 2.2. Traffic data
With all these policies being implemented simultaneously it is
very difficult to isolate their individual response and evaluate their Traffic data from Paris was provided by the Paris City Hall (Paris,
effectiveness. To overcome this, we evaluated trends in air pollution 2017) and comprised average vehicle flow from 7 a.m. to 9 p.m.
concentration in Paris and London seeking areas where the policy Monday to Friday and the ratio of vehicles between weekends and
mix was effective and areas where it was not. Linear trends using weekdays working hours. Annual Average Daily Flows (AADF, i.e.
the Theil-Sen estimator were calculated from 2005 to 2016 split in average number of vehicles per day) for roads in inner Paris and on
two: 2005e09; and 2010e16 for consistency with Font and Fuller the ring-road were calculated from the data provided. The pro-
(2016). These dates approximate the introduction of tighter Euro portion of each vehicle category in the fleet was available for given
standards for new cars and LGVs - Euro 4 (January 2006) and Euro 5 years (2005, 2010, 2012 and 2014) from Airparif based on the na-
(January 2011); 2010 is the mid-point. Trends in concentrations in tional inventory (CITEPA, 2017) corrected by local studies.
urban background and traffic sites were compared to establish AADFs were available from the Department for Transport (DfT)
whether changes were driven by the regional component or by for roads in London for different vehicle categories: cars & taxis,
traffic sources. Trends in roadside increments were also calculated motorcycles, buses & coaches, light good vehicles (LGVs) and heavy
to better evaluate the effectiveness of traffic-related measures in goods vehicles (HGVs). Traffic data was extracted for the nearest
reducing air pollutants concentrations. Linear-mixed effect models traffic counter (<1 km) from each monitoring site located on the
was applied to annual mean roadside increments to determine the same road. A total of 79 traffic counts locations were used. The
main traffic variables driving ambient trends and to rank them as a proportion of each vehicle category for London was calculated from
function of their importance as modulators of the air pollution the 79 roads for 2005, 2010, 2012 and 2014 (the same years as data
concentrations. was available for Paris).
 A. Font et al. / Environmental Pollution 249 (2019) 1e12 3

2.3. Trend calculations Possible collinearity between the explanatory variables was
assessed by means of pairwise Pearson correlation values, the
Linear trends in air pollutants concentrations, roadside in- conditional number (CN) and the maximum variance inflation
crements and traffic counts were calculated for two periods of factor (VIF). Thresholds of r > 0.70, CN > 30 and VIF>10 have been
time: 1 January 2005 to 31 December 2009; and 1 January 2010 to suggested to indicate possible collinearity (see references in
31 December 2016. Dormann et al., 2013). The majority of explanatory variables
The methodology to calculate trends was identical to Font and showed r < 0.70 with the exception of AADF Motorcycles - AADF
Fuller (2016) and a summary can be found in the Supplementary Heavy Vehicles (r ¼ 0.82); AADF Euro 4 light diesels - AADF Heavy
Information. Trends were expressed in absolute values (mg m
3 vehicles (r ¼ 0.8); and AADF Euro 4 light diesels - AADF Motorcycles
year
1) and as a percentage (% year
1). Trends in traffic increments (r ¼ 0.77) in the incPM2.5 model; and AADF Euro 5 light diesels -
(incNOX, incNO2, incPM10 and incPM2.5) were calculated by sub- AADF Euro V heavy vehicles (r ¼ 0.81); and AADF Euro 4 light
tracting a background concentration from each hourly roadside diesels - AADF Euro IV heavy vehicles (r ¼ 0.76) in the PM10 model.
measurement (“Lenschow approach”) (Lenschow, 2001). Note that CN and VIF were all below the 30 and 10 thresholds, indicating no
the approach may not fully apportion the impact of traffic emis- collinearity. The models for incPM2.5 and incPM10 were re-run
sions to background concentrations. However, the formation of without one of the correlated variables and regression co-
secondary air pollution generally takes place at time scales longer efficients compared to the base model. Regression coefficients did
than hourly and outside the city (Abdalmogith et al., 2006). At the not show a change of sign. Only the effect of AADF heavy vehicles in
urban spatial scale this is not thought to lead to a large bias. the incPM2.5 model increased when AADF motorcycles were
Trends were evaluated for each individual monitoring site in excluded. Details can be found in the Supplementary Information
Paris and London and then overall trends were calculated using S.4.
meta-analysis statistical tools. This is an effective methodology to
evaluate the variability of ambient responses to polices across a 3. Results
large urban area, accounting for individual and population-wide
variability among roads and other monitored locations as shown 3.1. Annual mean concentrations
previously by Font and Fuller (2016).
Trends in traffic were calculated as the slope resulting from the Throughout our study period, Paris and London measured
least-square linear model of AADF per year and were expressed in exceedances of the European Limit Value (ELV) for NO2 (40 mg m
3
units of Dvehicles day
1 year
1. In London, overall trends for AADF annual mean) across their networks. This limit value should have
were also calculated by means of the Random Effects (RE) model been met in 2010. Breaches took place in both background and
using the traffic counts near to the monitoring sites. roadside locations (Fig. 1). However, all background locations in
Paris attained the ELV in 2016. That was not the case for London; in
2.4. Linear-mixed effect models 2016 a few central background locations were still above the ELV.
Annual NO2 concentrations in roadside locations were slightly
Linear-mixed-effect models were built to identify the main higher in Paris (median annual NO2 across the network ranged
drivers of annual roadside increments concentrations in Paris and from 53 to 93 mg m
3 depending on the year) than in London
London. The annual roadside concentration for each pollutant (median across the network: 55e59 mg m
3). However, they were
(incNO2, incNOX, incPM10 and incPM2.5) at each monitoring site in quite similar in 2016 (median NO2 concentration across the road-
Paris and London between 2005 and 2016 (Dormann et al., 2013) side sites in the network was 59 and 58 mg m
3 in Paris and London,
was taken as outcome and a random effect was specified on the respectively). Notably some roads in London observed annual mean
monitoring site. All roadside sites in London were considered in the concentrations more than five times the ELV (i.e. WA7 and LB4)
model. For Paris, data availability meant that only “intra-muros” whereas the worst case in Paris was 2.7 times the ELV (i.e. AUT).
and for “Boulevard Pe
riphe

rique” mean roadside increments were However, almost all roadside locations in Paris and London
used. All sites in Paris and London were combined in a single model. exceeded the ELV every year for NO2 between 2005 and 2016
Several model formulations were built using different fixed ef- (Fig. 1).
fects: total traffic flow (AADF); flow for diesel cars and LGVs (named PM10 concentrations were below the annual limit value
light diesels); HGVs and buses & coaches (heavy vehicles) and (40 mg m
3) at all background locations in both Paris and London.
motorcycles. Also, the traffic flow for Euro 4 and Euro 5 light die- But some roadside locations in Paris still had problems in meeting
sels; Euro III, Euro IV and Euro V heavy vehicles; and Euro 2 and this limit; those located on the ring-road or next to major roads
Euro 3 motorcycles were also included as fixed effects in some of except for the BASCH site in inner Paris. PM10 in 2016 was exceeded
the model formulations. The presence of a Low Emission Zone (LEZ) only at A1 site. Roads in London did not exceed the annual ELV for
and city (Paris, London) were also included as categorical fixed PM10 during the study period except for LB4 in 2005 (Fig. 1).
effects. Supplementary Table 8 summarizes the model
formulations. 3.2. Traffic data: counts and Euro norm distribution
The best model for each pollutant was selected based on the
Conditional Akaike Information (cAIC); choosing the model with The traffic count on the monitored roads in London was slightly
the lowest cAIC value. To test the significance of the selected model, greater (~34,000 to 31,000 vehicles day
1 from 2005 to 2014)
a null model without the fixed effects was built and its cAIC value compared with Paris (~30,000 to 24,000 vehicles day
1). It should
computed. Ideally, cAIC of the selected model should be lower than be noted that the traffic flow for the Parisian ring-road was greater
that from the null. Also, the likelihood ratio test was done (~100,500 vehicles day
1) than other roads in Paris (Supplementary
comparing the selected and the null models using the anova test. Tables 3 and 4). The vehicle fleet in both cities was dominated by
The model was statistically different from the null model if p < 0.05. cars and taxis representing ~70e75% of the fleet; followed by LGVs
Linear-mixed effect models were computed using the ‘lmer’ func- (~15e16% in Paris and 12e15% in London); and HGVs (~6% and ~5%,
tion of the lme4/R package (version 1.1e17) (Bates et al., 2014). The respectively). The rest of the fleet (~7%) was dominated by motor-
95% confidence intervals for the fixed effects were calculated using cycles in Paris whereas it was split almost equally between mo-
the ‘confint’ function from the MASS/R package (version 7.3e50). torcycles and buses & coaches in London. Notably, the proportion of
4 A. Font et al. / Environmental Pollution 249 (2019) 1e12

Fig. 1. Annual mean NO2 vs annual PM10 concentrations at background and roadside locations in 2006, 2011 and 2016 in Paris and London. Annual Limit Values are marked with red
lines. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

buses in the Parisian fleet (~0.5e0.6% of the total) was less than one series. Most remaining Euro III HGVs after 2012 in London were
seventh of the share in London. adapted to meet Euro IV standards for PM emissions (63e93%) and
Diesel cars dominated the Parisian fleet in the whole study therefore permitted in the LEZ. Euro V was introduced in 2010 and
period and by 2015 they represented 74% of the cars driven in the by 2015 it had become a large part of the fleet, 40% in Paris and 51%
city. The share of diesel in London was lower but increased from in London.
~20% in 2005 to 43% in 2015. The number of diesel LGVs was slightly The Euro class distribution of buses & coaches in Paris was
larger in London (95e98%) than Paris (84e91%) (Supplementary similar to that of HGVs for the whole time period. In London, Euro
Fig. 20). III buses dominated the fleet up to 2012 when Euro V was the main
The distribution in Euro standard for each vehicle category can standard. Euro IV was introduced in 2006 and by 2008 reached 40%.
be found in Fig. 2. In 2005-09 diesel cars in Paris comprised Euro 3 Euro II buses represented ~40% in 2005 but were less than 10% from
(41e30%), Euro 2 (25e16%) and Euro 1 (18e7%). Euro 4 cars were 2008 onwards. Euro VI buses & coaches were introduced in 2013 in
introduced in 2006 and this was the main Euro class by 2009 Paris, a year before than London; and by 2015 its share was 20%
representing ~50%. By contrast, for the same period, Euro 3 domi- compared to 12% in London.
nated the diesel car fleet in London (~70% in 2005) and reflecting
the growth in the proportion of diesel cars (Supplementary section 3.3. Trends in air pollutants
S3.6) these were slowly replaced by Euro 4 after its implementation
in 2006. By 2009 Euro 4 represented 51% of the diesel cars in Nitrogen oxides and particulate matter in both Paris and London
London. In 2010e16 the distribution of Euro classes in the diesel car have generally improved with downward trends in both periods of
fleet was quite similar in both cities, with a dominance of Euro 4 time (Table 1). Both cities had downward trends in background NO2
(40e50%); and the replacement of pre-Euro 4 cars by Euro 5. By concentrations in 2005e09 (
2.1% year
1 and -1.4% year
1 in Paris
2016, Euro 4 and Euro 5 represented 80% of the car fleet in both and London, respectively) while roadside trends did not show any
cities. The distribution of LGVs by Euro class was quite similar to the significant change. In 2010-16 NO2 concentrations at background
car distribution in each city. sites in London decreased faster (
2.1% year
1) than Paris (
1.7%
At the start of the 2005e09 period both Paris and London had year
1). However, roads in Paris and London showed a similar trend
similar HGV Euro distribution with Euro II and III dominating the within uncertainties (
2.9 [-3.4,
2.4] and
2.3 [-2.9,
1.8] %
fleet. But at the end of the period the distribution was very different year
1, respectively).
in the two cities. Euro II HGVs were quickly replaced in London's Paris had no significant trend in PM10 during the first period of
fleet by Euro IV and had almost disappeared by 2008, when the LEZ time but had a fast downward trend of
4.4% year
1 in both
was introduced. Euro IV was introduced into the fleet of the two background and roadside locations during the second period.
cities but reached a higher share in London (49% by 2009) than London showed significant downward trends in PM10 concentra-
Paris (30%). By 2009 the presence of Euro II HGVs was still notable tions at background and roadside locations in both periods. How-
in Paris at 20%. By 2010 HGVs in both Paris and London were mainly ever, in the first period background locations observed a faster
Euro III (~40%) and IV (40% and 52% in Paris and London, respec- decrease (
3.7% year
1) than roadside sites (
3.0% year
1).
tively). The introduction of phase 4 of the LEZ in London induced a Downward trends in 2010e16 were similar at both background and
faster decrease in Euro III compared to Paris at the end of the time roadside (
2.9% year
1 and -3.0% year
1, respectively).
 A. Font et al. / Environmental Pollution 249 (2019) 1e12 5

Fig. 2. Distribution of Euro classes for diesel vehicles (cars, LGVs, HGVs and buses & coaches) for Paris and London.

Table 1
Overall percentage trends calculated by means of the random-effects linear model for NOX, NO2, PM10 and PM2.5, for the periods 2005e2009 and 2010e2016 in London and
Paris by station type. Trends in roadside increments are also reported. Brackets denote 95% confidence intervals. *** significant at the 0.001 level; ** significant at the 0.01 level;
* significant at the 0.05 level; þ significant at the 0.1 level; (blank) not statistically significant.

City Pollutant 2005e09 2010e16

Background trend (% Roadside trend (% Roadside increment (% Background trend (% Roadside trend (% Roadside increment (%
year
1) year
1) year
1) year
1) year
1) year
1)

Paris NOX
1.7 [
2.5,
0.9]***
3.1 [
5.7,
0.5]*
3.5 [
7.6, 0.6]þ
1.5 [
1.9,
1.1]***
1.9 [
2.5,
1.4]***
2.7 [
3.5,
1.8]***
NO2
2.1 [
2.82,
1.3]***
0.5 [
2.4, 1.5] 0.7 [
2.9, 4.2]
1.7 [
2.0,
1.4]***
2.9 [
3.4,
2.4]***
5.5 [
7.0,
4.0]***
PM10 0.2 [
1.0, 1.5]
0.03 [
1.8, 1.8]
0.3 [
2.3, 1.7]
4.3 [
5.0,
3.6]***
4.3 [
4.7,
3.9]***
5.4 [
6.4,
4.3]***
PM2.5
4.2 [
7.5,
1.0]*




5.2 [
6.2,
4.1]***
6.6 [
7.3,
5.9]***
10.5 [
12.6,
8.4]***
London NOX
2.1 [
2.6,
1.5]***
1.3 [
2.0,
0.7]*** 1.1 [0.1, 2.2]*
1.6 [
2.2,
1.0]***
1.5 [
2.1,
0.8]***
1.7 [
2.7,
0.8]***
NO2
1.4 [
2.1,
0.7]***
0.2 [
0.9, 0.5] 10.4 [7.9, 12.8]***
2.1 [
2.7,
1.6]***
2.3 [
2.9,
1.8]***
5.0 [
3.1,
6.9]***
PM10
3.7 [
4.7,
2.7]***
3.0 [
3.6,
2.4]***
3.0 [
5.9,
0.1]*
2.9 [
3.9,
2.0]***
3.0 [
3.6,
2.4]***
8.4 [
4.4, 12.5]***
PM2.5


5.1 [
8.6,
1.6]**


4.4 [
5.3,
3.5]***
3.7 [
5.2,
2.1]***
2.6 [
12.5, 7.3]

PM2.5 concentrations at background locations in Paris decreased observed in background (
4.4% year
1).
at
4.2% year
1 in 2005e09 (note that only two sites were avail- Overall trends in roadside NO2 increments changed sign from
able; Supplementary Fig. 6) and faster in 2010e16 at
5.2% year
1. the 2005e09 period, which had an upward trend in both cities, to a
PM2.5 along roads was only available in the second period and significant downward trend in 2010e16 (
6% year
1 and -5%
showed a decrease of
6.6% year
1, faster that their background year
1 in Paris and in London, respectively) (Table 1). Trends in
counterparts. London had downward trends in both periods. roadside PM10 increment were flat in 2005e09 in both cities and
However, downward trends in roadside locations were slower in then showed a downward trend from 2010 (
5% year
1 in Paris
2010e16 (
3.7% year
1) than in 2005e09 (
5.1% year
1). Trends and
8% year
1 in London). Trends in roadside PM2.5 increment
along roads in 2010e16 were slower (
3.7% year
1) than those were only available in the second period. Paris observed a
6 A. Font et al. / Environmental Pollution 249 (2019) 1e12

significant downward trend (
11% year
1) and London showed a fastest increase (3.2 mg m
3 year
1). In 2010e16 all roads showed a
downward but non-significant trend (
2.6% year
1) (Table 1). It significant downward trend for all pollutants (Fig. 4) although the
should be highlighted that trends in roadside PM2.5 increment in rate of response varied. The trend in incNO2 and incPM10 in
Paris was estimated from two monitoring sites. 2010e16 in Boulevard Pe
riphe
rique Porte d’Auteuil was twice as
When looking at the behaviour of individual roads, we can see fast (
4.3 and
1.7 mg m
3 year
1, respectively) as the overall trend
considerable heterogeneity. Trends in roadside incNO2 in ^Ile-de- (
1.9 and
0.9 mg m
3 year
1, respectively) (Fig. 4A, C).
France showed an increase in 2005e09 at almost all roads except Compared with Paris, London roads had a greater variability of
for Quai des Ce
lestins (Fig. 3A) that showed a significant downward responses for all pollutants in the two periods of time but especially
trend. This response was due to the change of the layout of the road in 2010e16. Despite the overall downward trend in incNO2 and
with a construction of a bicycle lane between the monitoring site incPM10 in 2010e16, there were some roads with significant up-
and the road. Boulevard Pe
riphe
rique Porte d’Auteuil observed the ward trends (Fig. 4B, D). Trends in incPM2.5 also had a variability of

Fig. 3. Forest plots for the trends in roadside increments in NO2 (incNO2) and PM10 (incPM10) for Paris and London for 2005e09.
 A. Font et al. / Environmental Pollution 249 (2019) 1e12 7

Fig. 4. Forest plots for the trends in roadside increments in NO2 (incNO2), PM10 (incPM10), and PM2.5 (incPM2.5) for Paris and London for 2010e16.
8 A. Font et al. / Environmental Pollution 249 (2019) 1e12

response; some roads had fast downward trends (Greenwich e and 5 heavy vehicles; and Euro 2 motorcycles (0.7, 1.2 and
A206 Burrage Grove and Westminster-Marylebone Road) and 3.5 mg m
3 1000 vehicles
1, respectively) whereas AADF Euro 5
others had significant upward trends (e.g. Kensington and Chelsea light diesels was associated with a decrease in incPM10
e Cromwell Road) (Fig. 4F). (
0.4 mg m
3 1000 vehicles
1). Roadside PM2.5 increments only
showed two traffic parameters with significant coefficients: AADF
3.4. Traffic: counts and trends motorcycles (1.2 mg m
3 1000 vehicles
1) and AADF Euro 5 light
diesels (
0.3 mg m
3 1000 vehicles
1).
The total number of vehicles decreased during the study period
in the two cities. Paris observed similar rates in both periods of time 4. Discussion
(around
2% year
1) while London observed a faster decrease in
2005e09 (
1.4% year
1) than in 2010e16 (
0.2% year
1) 4.1. Nitrogen dioxide (NO2)
(Supplementary Table 5). Trends on the ring-road in Paris showed a
different behaviour, with a non-significant upward trend in the 4.1.1. Trends in roadside increments in 2005e09
2005e09 and a modest downward trend in 2010e16 (
0.3% year
1, Roadside locations in Paris and in London did not benefit from
significant at p < 0.1) (Supplementary Table 6). the downward trends observed in background locations in
Only London had data to calculate trends in each vehicle cate- 2005e09. Trends in incNO2 showed an upward trend in both cities
gory. Here, the downward trend in the total number of vehicles was (the lack of statistically significant trend in Paris was due to the fast-
largely explained by a downward trend in the number of cars & downward trend due to the road layout changes at Quai des
taxis (
1.5 and
0.3% year
1 in the first and second period, Ce
lestins site. When this is removed from the analysis, the overall
respectively) and a change in the number of HGVs (
0.8 and
0.6% trend in incNO2 was an increase of 2.5% year
1**).
year
1). The number of LGVs increased in both periods but not at a Counterintuitively, the increase in incNO2 took place when the
significant rate. Buses & coaches increased in 2005e09 (1.7% number of vehicles in the urban centres decreased fastest,
year
1) but decreased after 2010 (
0.6% year
1). Motorcycles around
2.1% year
1 in Paris and
1.5% year
1 in London. However,
increased in both time periods but remarkably in the second period the proportion of diesel vehicles increased: from 56% (2005) to 74%
at a rate of 11% year
1 (Supplementary Table 7). (2009) of cars in Paris; and 20%e33% of in London (Supplementary
Fig. 20). The linear-mixed effect model indicated that AADF of light
3.5. Influence of traffic variables to roadside increment diesel vehicles (cars þ LGVs) was associated with an increase in
concentrations incNO2. The mean number of light diesel vehicles along roads in
Paris and London increased from ~12,200 in 2005 to ~15,200 in
Table 2 indicates the selected linear-mixed effect model for each 2009, representing an increase of 1.8 mg m
3 in incNO2. The other
pollutant and the statistical parameters. The cAIC values and the R2 traffic parameters associated with higher incNO2 with significant
for each model including the null model can be found in the change between 2005 and 2009 were Euro IV heavy vehicles, that
Supplementary Table 9; and the formulation for the selected model were introduced in the fleet with a mean of 1200 vehicles by 2009,
for each pollutant is shown in Supplementary Table 10. adding ~3.5 mg m
3.
The traffic parameters with greatest impact on roadside NO2 In 2005e09, Euro 4 cars and LGVs were introduced into the fleet
increments were AADF Euro 4 heavy vehicles (2.9 mg m
3 every of the two cities and was the main standard by 2009 (Fig. 2).
1000 vehicles) > AADF motorcycles (1.8 mg m
3 1000 vehi- Despite the stringent emission standards for NOX (the pollutant
cles
1) > AADF Euro 3 heavy vehicles (1.7 mg m
3 1000 vehi- regulated in vehicle performance tests) in Euro 4 diesel cars and
cles
1) > AADF light diesel vehicles (0.6 mg m
3 1000 vehicles
1). LGVs, evidence has shown that under real driving conditions
AADF Euro 5 heavy good vehicles lead to a decrease on incNO2 emissions did not align with improved performance in approval
levels at a rate of
4.1 mg m
3 1000 vehicles
1. tests (Carslaw and Rhys-Tyler, 2013; Carslaw et al., 2016; Sjo €din
The linear-mixed effect model for incNOX did not separate the et al., 2017). NO emitted in urban areas would have been trans-
different Euro norms for heavy vehicles but did it for light diesels. formed to NO2 by O3 reaction. Furthermore, there is evidence that
Also, the AADF of motorcycles and heavy vehicles were associated primary NO2 emissions from light diesel vehicles measured in
with an increase in incNOX (4.8 and 3.5 mg m
3 1000 vehicles
1) London increased from 16% (Euro 3) to 28% (Euro 4) of total NOX
and AADF Euro 5 light diesels was associated with a decrease (Carslaw et al., 2016). Slightly higher figures were reported in
(
2.4 mg m
3 1000 vehicles
1). Go€teborg for Euro 4 (~35%) (Sjo €din et al., 2017). The high NO2
Roadside increments in PM10 were influenced by AADF Euro 3 emissions from Euro 4 were due to the vehicle after-treatment

Table 2
Statistical parameters (intercept and regression coefficients) of the selected linear-mixed effect models for each pollutant. Values denote the 95% confidence interval. Units for
fixed effects are expressed per 1000 vehicles. Bold numbers indicate significant at 95% level.

incNOX incNO2 incPM10 incPM2.5

Best model formulation #5 #6 #8 #5

Intercept 93.8 [71.5, 116.5] 17.4 [8.9, 25.8] 4.2 [2.8, 5.7] 1.2 [-0.5, 2.8]
AADF light diesels 0.6 [0.3, 1.0]
AADF heavy vehicles 3.5 [1.1, 5.8] 0.4 [-0.1, 1.0]
AADF motorcycles 4.8 [1.5, 8.3] 1.8 [0.2, 3.3] 1.2 [0.9, 1.6]
AADF Euro 4 light diesels 0.3 [-0.2, 0.8] 0.1 [-0.03, 0.3] 0.0 [-0.2, 0.1]
AADF Euro 5 light diesels ¡2.4 [-3.0, -1.9] ¡0.4 [-0.5, -0.2] ¡0.3 [-0.5, -0.2]
AADF Euro III heavy vehicles 1.7 [0.1, 3.3] 0.7 [0.1, 1.2]
AADF Euro IV heavy vehicles 2.9 [1.6, 4.3]
0.5 [-1.3, 0.3]
AADF Euro V heavy vehicles ¡4.1 [-5.6, -2.6] 1.2 [0.6, 1.8]
AADF Euro 2 motorcycles 3.5 [1.9, 5.1]
AADF Euro 3 motorcycles 0.2 [-0.6, 0.9]
 A. Font et al. / Environmental Pollution 249 (2019) 1e12 9

technologies to reduce carbon monoxide and hydrocarbons; and to between 2 and 193 years (average of 21). The background sites in
aid control of PM diesel emissions (Carslaw et al., 2016). Conversely, London that exceeded might still need 6 years (ranging from 1 to 23
on-road emissions from Euro IV heavy vehicles showed a reduction years) to comply at current rates. However, it should be remem-
in the primary NO2 fraction (Carslaw et al., 2016) despite an in- bered that these calculations do not include new policies such as
crease in NOX emissions (Carslaw and Rhys-Tyler, 2013; Sjo € din et al., the introduction of Euro 6 (a-d)/VI in the fleet. Grice et al. (2009)
2017). estimated that the annual mean concentration of NO2 in London
The greater increase in incNO2 in 2005e09 observed on the is expected to remain above the limit value until 2020 or beyond
Paris ring-road (Boulevard Pe
riphe

rique Porte d’Auteuil) compared even with new Euro VI standards to address the emissions from
to the inner roads in Paris, and based on the results of the linear heavy duty vehicles. Despite tightening NOX emission limits on new
mixed-effect model for incNO2, is explained by the larger increase Euro 6 vehicles and the introduction of the real-driving emissions
in light diesel vehicles (14% compared with 5.5%, respectively) and tests in Euro 6d-temp, the allowed conformity factors will delay the
an increase of ~2300 Euro IV heavy vehicles compared with ~600 in widespread adoption of available emission control technologies
inner Paris. (Degraeuwe et al., 2017) and alternative measures such as the
reduction of diesel cars and move to a gasoline urban fleet would be
4.1.2. Trends in roadside increments 2010-16 a more effective route to achieve compliance with the EU standards.
In contrast to 2005e09, there were clear downward trends in The roads that had faster downward trends in 2010e16 in NO2 in
incNO2 in 2010e16 in both cities at 5e6% year
1. The linear-mixed London were those with the greatest annual mean NO2 concen-
effect model indicated that Euro V heavy vehicles were responsible tration in 2010 (CD9, LB4, WA7 and WA8) (Supplementary Fig. 21).
for reductions in incNO2. This agrees with real-world emission tests In the case of WA7 and WA8 an intervention program fitting Euro III
from Euro V HGVs that show a reduction of 22e85% in primary NO2 buses with Selective Catalyst Converters (SCRs) in 2013 successfully
emissions compared with Euro II/III standards (Carslaw et al., 2016; reduced roadside NOX and NO2 concentrations by 20% (Barratt and
Sjo€din et al., 2017). By 2015, Euro V dominated the HGV and bus & Carslaw, 2014). Despite the success of this intervention, it cannot be
coach fleets in both cities (Fig. 2). Therefore, it is likely that the repeated (although new intervention programmes can be
downward trends in incNO2 concentrations in 2010e16 in the two deployed). The overall downward trend at these locations
cities reflected the reduction of NO2 emissions from Euro V heavy decreased when the step-change was viewed over the 2010e16
vehicles. Another factor that might have hastened the decrease in period compared with 2010e14 in Font and Fuller (2016). Further
incNO2 in 2010e16 is the increase of light diesel vehicles, most policies will clearly be required to reduce NO2 ambient levels.
likely explained by the negative coefficient in incNOX (Table 2). The emission performance of buses & coaches will be particu-
There was a faster downward trend in incNO2 on the ring-road larly relevant for London and especially on those roads where they
site compared with inner Paris roads. This was consistent with the represent an important fraction of heavy vehicles. TfL buses within
greater increase of Euro V heavy vehicles (380%) on the ring-road the London's LEZ should comply to Euro IV standards from
compared with inner roads (332%). December 2017 (Holman et al., 2015). On-road tests of Euro IV (Liu
London had a faster HGV fleet turnover than Paris due to the et al., 2011) and Euro V buses (Zhang et al., 2014) found an
introduction of the LEZ in 2008, as seen in Fig. 2 and as shown improved performance in NOX emissions compared to previous
separately by Ellison et al. (2013). Pre-Euro III HGVs were removed emission standards but they still exceeded the regulated test limit.
from the fleet and replaced by Euro IV (up to 2010) and then by According to our model results, Euro IV heavy vehicles were not
Euro V. The tightening of the LEZ in 2012 led to the remaining Euro associated with a decrease in incNO2 concentrations either. Zhang
III HGVs being retro-fitted to meet Euro IV PM standards (e.g. fitting et al. (2014) suggested that hybrid and natural gas buses as the
a diesel particle filter - DPF). However, the LEZ did not produce alternatives to control NOX emissions however others (Howarth,
faster reductions in incNO2 concentrations when compared with 2014) have suggested large leakages from the use of methane as a
Paris. The LEZ in London was designed to reduce PM emissions but road fuel could be harmful for climate change.
was also predicted to reduce NOX concentrations in London by 18% It is also important to note that some roads in London that
(Cloke et al., 2000) based on assumed real-world emissions exceed NO2 ELV concentrations had upward trends in incNO2
declining in line with Euro standards. The linear-mixed effect (Fig. 4B). Clearly policies were not being successful along these
model suggests that reductions in incNO2 would be achieved roads. Except for Islington e Holloway Road (IS2) and Hackney e
tightening standards of heavy vehicles to Euro V. Old Street (HK6), the majority of these roads are located far from
The Parisian fleet still had Euro II and III HGVs in 2015 ac- the city centre (>10 km) (Supplementary Fig. 21). In 2010e14, roads
counting for 30% of the HGVs. A LEZ was introduced in Paris in with increased incNO2 were associated with an increase in the
September 2015 banning pre-Euro III HGVs. Based on the London's number of heavy vehicles (buses and HGVs) (Font and Fuller, 2016).
experience replacing pre-Euro III HGVs is not expected to provide
rapid NO2 reductions if these are solely replaced by Euro IV as
shown by the trends in incNO2 in 2005e09 in London and 4.2. Particulate matter (PM10 and PM2.5)
confirmed by the linear-mixed effect model. The LEZ in Paris also
bans diesel cars and LGVs (pre-Euro 3) and motorcycles (pre-Euro Trends in PM10 at the background and the roadside in Paris were
2). However, on-road primary NO2 from pre-Euro 3 cars and LGVs similar in magnitude for each period; however, trends in roadside
was less than Euro 4 and 5 (Carslaw et al., 2016). The effectiveness increments showed a faster decrease. The same pattern was
of the LEZ would probably be maximised by tightening the policy to observed in London. The traffic increment represented on average
also exclude Euro 4 and/or Euro 5 LGVs; and Euro pre-V HGVs. ~35% and 18% of the total PM10 concentration measured at the
roadside locations in Paris and London, respectively. Therefore,
4.1.3. Future directions and considerations trends at roadside locations would have been largely determined
Despite the recent downward trends, both Paris and London are by the downward trends in the urban background. To be effective,
still a long way away from attaining the NO2 ELV which was set for the reduction of urban PM10 concentrations should also include
2010. To place current rate of progress in context, based on trends wider urban and regional policies such as tackling urban wood
between 2010 and 2016, roads in Paris will still need between 4 and burning and secondary PM precursors, including vehicle emissions
20 years (average of 10) to attain the ELV. For London's roads this is from outside the city boundaries.
10 A. Font et al. / Environmental Pollution 249 (2019) 1e12

4.2.1. Trends in roadside increments 2005-09 Parisian roads with PM2.5 data were greater than those observed in
Policies reducing vehicular PM emissions were successful in background locations indicating the success of traffic-related pol-
London since 2005e09 and in Paris since 2010e16 (Table 2) with icies (Fig. 4E); however, it was the opposite in London where trends
downward trends in the traffic increment. in roadside incPM2.5 in 2010e16 showed a non-significant down-
Paris observed flat non-significant trends in 2005e09 while ward trend (Table 1). This behaviour contrasted with earlier anal-
London observed significant downward trends. The linear-mixed ysis of 2010e14 when most roads in London observed significant
effect model suggested that the downward trend in London was downward trends (Font and Fuller, 2016). Trends in incPM2.5 were
associated with the replacement of Euro III HGVs with Euro IV. not monotonic in 2010e16 and the consistent downward trend
Although the model for incPM10 did not result in a statistically observed in 2010e14 was broken by an increase in the roadside
significant coefficient for Euro IV HGVs, the central estimate PM2.5 concentration in 2015e16 (Supplementary Fig. 19). This later
showed a negative coefficient (Table 2). Moreover, less Euro III increase was more prominent on some roads (e.g. CD1: Camden e
heavy vehicles was associated with less incPM10. The benefit in Swiss Cottage) than others (e.g. MY7: Westminster e Marylebone
replacing Euro III HGVs in Paris may have been offset by the large Road) resulting in a non-significant downward trend when
proportion of motorcycles. considering the period 2010e16. Some roads in London had slower
Trends in incPM2.5 were only available at the roadside locations trends in PM2.5 than that at the background (KC1) (Supplementary
in London. It showed a faster decrease (
5.1% year
1) than incPM10 Fig. 22).
(
3.0% year
1). The linear-mixed effect model for incPM2.5 identified Euro 5
light diesels as the vehicle category associated with reducing
4.2.2. Trends in roadside increments 2010-16 roadside PM2.5; and motorcycles with increasing factors (Table 2).
Paris and London observed downward trends in incPM10 in The number of motorcycles in London has increased rapidly at a
2010e16. According to the linear-mixed effect model for incPM10 rate close to 11% year
1 (Supplementary Table 7). This increase
(Table 2), the introduction of Euro 5 light diesels significantly might be associated with the proliferation of motorcycles based
reduced roadside PM10 concentrations. This is due to the exhaust deliveries in recent years (e.g. home meal-delivery) or the perma-
diesel particle filters (DPFs) (Fiebig et al., 2014) that were effectively nent opening of London's bus lanes to motorcycles in 2011 (TfL,
compulsory in Euro 5 cars and LGVs (2011). DPFs in heavy vehicles 2011). This is an important issue since Euro standards for motor-
were first introduced in Euro IV (2009). The slower decrease in cycles do not regulate PM emissions (except for quads). Despite
incPM10 in Paris compared to London might be due to the greater small engine size, motorcycles emissions tests carried out by Pham
proportion of motorcycles that offset the benefits of Euro 5 light et al. (2013) showed that they emitted more PM than petrol light
diesel vehicles. duty gasoline vehicles reported in Robert et al. (2007) with PM
Surprisingly, the model gave non-significant factor for Euro IV emissions from motorcycles that were much greater even than
heavy vehicles and positive factors for Euro V, despite the DPFs. A non-catalyst petrol vehicles (model years 1965e2003) for the same
possible explanation is that the reduction of PM10 emissions from distance. Motorcycles emit significant quantities of primary organic
tail-pipe (in the fine fraction) might have been counteracted by aerosols, aromatic volatile carbon components and also are
increased non-exhaust particle emissions (resuspension, brake- responsible for the production of secondary organic aerosols (Platt
wear, tyre-wear) (dominant in the coarse fraction). An increase in et al., 2014). The poor combustion in motorcycles in terms of air fuel
the incPM10 from non-exhaust emissions might be influenced by an ratio is often responsible for high unburned hydrocarbons and
increase in the traffic flow, especially that of heavy vehicles. An particle emission levels, with submicron particles dominating the
increase in kerbside coarse particle concentrations were observed exhaust (Yang et al., 2005; Chien and Huang, 2010). The increase in
in a central location in London following a higher number of buses the number of motorcycles in London might therefore have led to
with increased vehicle weight associated with larger resuspension an increase in fine particle emissions, offsetting the expected
rates (Carslaw et al., 2006). However, heavy vehicles, both HGVs benefits of DPFs on other vehicle classes.
and buses & coaches, showed a decreasing trend in 2010e16 in The Parisian fleet has a larger share of motorcycles compared
London (Supplementary Table 7). Despite the overall downward with London. The relationship between motorcycles and PM2.5
trend in incPM10 observed across London's roads in 2010e16 concentrations in London therefore has clear implications for Paris.
(Table 1), there were some suburban roads that had upwards trends The two Parisian sites that observed a PM2.5 decrease in 2010e16
(Fig. 4D). With one exception, the eight roads with upwards trends are located on the ring-road and therefore had a lower share of
in incPM10 were suburban (distance from the city centre > 10 km) motorcycles than inner-Paris roads, which would be consistent
(Supplementary Fig. 21). The central London site (City of London e with the evidence here of their large PM2.5 emissions.
Upper Thames Street) was affected by nearby construction dust in
2016. All other central roads measured a decrease. This indicates 5. Conclusions
that the control on vehicular PM10 emissions did not have the same
response everywhere. Unfortunately, most roads with positive This paper evaluated recent trends (2005e2016) in traffic
incPM10 trends did not have collocated PM2.5 measurements related pollutants (NO2, NOX, PM10 and PM2.5) in Europe's two
(Fig. 4D, F; Supplementary Fig. 23) to confirm if an increase in the megacities, Paris and London. Monitoring in two different cities
coarse fraction was the cause. The plots relating trends in incPM10 allowed us to evaluate the success of policies (Euro norm, vehicle
vs trends in AADF for different vehicle categories in London traffic flow and Low Emission Zone) in a wider perspective. Linear-
(Supplementary Fig. 24) indicate that sites in outer London with mixed effect models provided information of which traffic variables
positive trends had an increase in LGVs. That appears to contradict drove the trend in roadside increment concentrations. The main
the results from the linear mixed effect model. However, the light results were:
diesel vehicle category in the model was not LGV specific and
included both cars and LGVs; and also the Euro norm. It might be - Light diesel vehicles, motorcycles and Euro III and IV heavy ve-
the case that these locations with upward incPM10 trends might hicles were associated with increased traffic NO2 concentrations
have older LGVs and the reduced PM10 traffic concentrations from in 2005e09 in both cities.
Euro 5 diesel cars might be offset with the light diesel vehicle count. - The downward trend in traffic NO2 concentrations in 2010e16 in
Decreasing trends in incPM2.5 concentrations on the two both Paris and London roads was associated with the
 A. Font et al. / Environmental Pollution 249 (2019) 1e12 11

introduction of Euro V heavy vehicles. The introduction of Euro Airparif, 2016. Inventaire re
gional des e
missions en Ile-de-France Anne
e de

fe
re
rence 2012 - e
le
ments synthe
tiques Edition mai 2016.
5 Euro norm in cars and light good vehicles did not result in a
Amato, F., Favez, O., Pandolfi, M., Alastuey, A., Querol, X., Moukhtar, S., Sciare, J.,
clear decrease in traffic NO2 in line with the evidence on the 2016. Traffic induced particle resuspension in Paris: Emission factors and source
real-world emissions tests (dieselgate) (Degraeuwe and Weiss, contributions. Atmos. Environ. 129 (JANUARY), 114e124. https://doi.org/10.
2017). 1016/j.atmosenv.2016.01.022.
Barratt, B., Carslaw, D.C., 2014. Impacts of the bus retrofit programme on NO2
- London observed a greater heterogeneity of behaviour with concentrations along Putney High Street. Environ. Res. Group, Rep. Prep. Lon-
some roads displaying an upward trend in incNO2 in 2010e16. don Brgh. Wandsworth.
Roads with greatest concentrations were improving; those with Bates, D., Ma €chler, M., Bolker, B., Walker, S., 2014. Fitting Linear Mixed-Effects
Models using lme4 (1). https://doi.org/10.18637/jss.v067.i01.
increasing NO2 had one to two times the annual limit value in Bessagnet, B., Hodzic, A., Blanchard, O., Lattuati, M., Le Bihan, O., Marfaing, H.,
2010. Rouïl, L., 2005. Origin of particulate matter pollution episodes in wintertime
- An overall downward trend in traffic increment of PM10 was over the Paris Basin. Atmos. Environ. 39, 6159e6174. https://doi.org/10.1016/j.
atmosenv.2005.06.053.
observed in London since 2005; and since 2010 in Paris. Notably, Carslaw, D.C., Rhys-Tyler, G., 2013. New insights from comprehensive on-road
this was not observed on all London roads. measurements of NOx, NO2 and NH3 from vehicle emission remote sensing
- The fast replacement of Euro III HGVs in London, partly induced in London, UK. Atmos. Environ. 81, 339e347. https://doi.org/10.1016/j.atmosenv.
2013.09.026.
by the introduction of the LEZ, explained the earlier downward
Carslaw, D.C., Ropkins, K., Bell, M.C., 2006. Change-point detection of gaseous and
trend in incPM10 when compared with Paris. The greater pro- particulate traffic-related pollutants at a roadside location. Environ. Sci. Tech-
portion of motorcycles in Paris might have offset the benefit of nol. 40, 6912e6918. https://doi.org/10.1021/es060543u.
Carslaw, D.C., Murrells, T.P., Andersson, J., Keenan, M., 2016. Have vehicle emissions
replacing Euro III HGVs.
of primary NO 2 peaked? Faraday Discuss 00, 1e16. https://doi.org/10.1039/
- Euro 5 light diesel vehicles were responsible for downward C5FD00162E.
trends in incPM10 since 2010. Chien, S.M., Huang, Y.J., 2010. Sizes and polycyclic aromatic hydrocarbon compo-
- Increment PM2.5 concentrations had a statically significant sition distributions of nano, ultrafine, fine, and coarse particulates emitted from
a four-stroke motorcycle. J. Environ. Sci. Heal. - Part A Toxic/Hazardous Subst.
downward trend in Paris in 2010e16 which was associated with Environ. Eng. 45, 1768e1774. https://doi.org/10.1080/10934529.2010.513289.
Euro 5 light diesel vehicles. London did not observe such clear CITEPA, 2017. Organisation et me
thodes des inventaires nationaux des e
missions
trend due to the increase of the motorcycle flow. atmosphe
roiques en France. Rapport OMINEA e 14e me e
dition 1365.
Cloke, J., Cox, J.A., Hickman, A.J., Ellis, S.D., Ingrey, M.J., Mtru, K.B., Report, T.R.L.,
2000. A low emission zone for London Prepared for. Westminster City Council.
The success of current and future policies relies on the real- DEFRA (Defra and the Devolved Administrations), 2010. Assessment of UK AURN
world emissions of new vehicles being less than the ones they Particulate Matter Monitoring Equipment against the January 2010 guide to
demonstration of Equivalence. AGG04003328/BV/AQ/DH/2658/V3, (December).
replace. It is clear from comparing the traffic composition between Degraeuwe, B., Weiss, M., 2017. Does the New European Driving Cycle (NEDC) really
the two cities that the London's LEZ successfully replaced the fail to capture the NOXemissions of diesel cars in Europe? Environ. Pollut. 222,
intended vehicle types. However, it did not provide the expected 234e241. https://doi.org/10.1016/j.envpol.2016.12.050.
Degraeuwe, B., Thunis, P., Clappier, A., Weiss, M., Lefebvre, W., Janssen, S.,
impact on ambient pollutant concentrations, especially in NO2. The
Vranckx, S., 2017. Impact of passenger car NOXemissions on urban NO2pollu-
responsibility for the transference of this success into ambient tion e Scenario analysis for 8 European cities. Atmos. Environ. 171, 330e337.
concentrations must therefore rest with the vehicle manufacturers https://doi.org/10.1016/j.atmosenv.2017.10.040.
Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carre
, G., Marque
z, J.R.G.,
that have previously engineered differences between type-
Gruber, B., Lafourcade, B., Leit~ao, P.J., Münkemüller, T., Mcclean, C., Osborne, P.E.,
approval tests and real-world emissions. Euro 6/VI vehicles will Reineking, B., Schro € der, B., Skidmore, A.K., Zurell, D., Lautenbach, S., 2013.
include on-road testing as part of their standard emission testing Collinearity: A review of methods to deal with it and a simulation study eval-
and an improvement in ambient air quality in urban areas is uating their performance. Ecography 36, 027e046. https://doi.org/10.1111/j.
1600-0587.2012.07348.x.
therefore expected. However, results from this study highlights that Ellison, R.B., Greaves, S.P., Hensher, D.A., 2013. Five years of London's low emission
motorcycles may be causing hotspots of air pollution in European zone: Effects on vehicle fleet composition and air quality. Transport. Res.
cities. A continued increase in the number of motorcycles and the Transport Environ. 23, 25e33. https://doi.org/10.1016/j.trd.2013.03.010.
Favez, O.F., Petit, J.-E., Bessagnet, B., Fre
de

rik, M., Chiappini, L., Lemeur, S., Leoz-
diesel car growth, especially light goods vehicles, could confound Garziandia, E., 2012. Main properties and origins of winter PM10 pollution
policies to reduce NO2 and PM concentrations. events in France Caracte
ristiques et origines principales des e
pisodes de
pollution hivernaux aux PM 10 en France Main properties and origins of winter
PM 10 pollution events in France. Poll. Atmos., (November 2015).
Acknowledgments Fiebig, M., Wiartalla, A., Holderbaum, B., Kiesow, S., 2014. Particulate emissions
from diesel engines: correlation between engine technology and emissions.
This research was funded by the Observatoire de l’air en ^Ile-de- J. Occup. Med. Toxicol. 9, 6. https://doi.org/10.1186/1745-6673-9-6.
Font, A., Fuller, G.W., 2016. Did policies to abate atmospheric emissions from traffic
France (Airparif) France. We would like to thank Gregor Stewart at
have a positive effect in London? Environ. Pollut. 1e12. https://doi.org/10.1016/j.
King's College London and Yvonne Brown at Transport for London envpol.2016.07.026.
for providing the distribution of Euro standards data for London. GLA (Greater London Authority), 2017. London Atmospheric Emissions Inventory
[WWW Document]. URL. https://data.london.gov.uk/dataset/london-
We would like to thank the London local authorities, Business
atmospheric-emissions-inventory-2013.
Improvement Districts and DEFRA (United Kingdom) for funding Green, D.C., Fuller, G.W., Baker, T., 2009. Development and validation of the volatile
measurement programmes in London. Part of AF and GWF salaries correction model for PM10-An empirical method for adjusting TEOM mea-
for this project were covered by the European Climate Foundation surements for their loss of volatile particulate matter. Atmos. Environ. 43 (13),
2132e2141. https://doi.org/10.1016/j.atmosenv.2009.01.024.
who also funded open access publication. Grice, S., Stedman, J., Kent, A., Hobson, M., Norris, J., Abbott, J., Cooke, S., 2009.
Recent trends and projections of primary NO2 emissions in Europe. Atmos.
Appendix A. Supplementary data Environ. 43, 2154e2167. https://doi.org/10.1016/j.atmosenv.2009.01.019.
Guerreiro, C., Gonzalez Ortiz, A., de Leeuw, F., Viana, M., Horalek, J., 2016. Air quality
in Europe - 2016 report. https://doi.org/10.2800/413142.
Supplementary data to this article can be found online at Holman, C., Harrison, R., Querol, X., 2015. Review of the efficacy of low emission
https://doi.org/10.1016/j.envpol.2019.01.040. zones to improve urban air quality in European cities. Atmos. Environ. 111,
161e169. https://doi.org/10.1016/j.atmosenv.2015.04.009.
Howarth, R.W., 2014. A bridge to nowhere: methane emissions and the greenhouse
References gas footprint of natural gas. Energy Sci. Eng. 2, 47e60. https://doi.org/10.1002/
ese3.35.
Abdalmogith, S.S., Harrison, R.M., Derwent, R.G., 2006. Particulate sulphate and Lenschow, P., 2001. Some ideas about the sources of PM10. Atmos. Environ. 35,
nitrate in Southern England and Northern Ireland during 2002/3 and its for- 23e33. https://doi.org/10.1016/S1352-2310(01)00122-4.
mation in a photochemical trajectory model. Sci. Total Environ. 368, 769e780. Liu, Z., Ge, Y., Johnson, K.C., Shah, A.N., Tan, J., Wang, C., Yu, L., 2011. Real-world
https://doi.org/10.1016/j.scitotenv.2006.02.047. operation conditions and on-road emissions of Beijing diesel buses measured
12 A. Font et al. / Environmental Pollution 249 (2019) 1e12

by using portable emission measurement system and electric low-pressure vehicles. J. Air Waste Manag. Assoc. 57, 1414e1428. https://doi.org/10.3155/
impactor. Sci. Total Environ. 409, 1476e1480. https://doi.org/10.1016/j. 1047-3289.57.12.1414.
scitotenv.2010.12.042.

Rouïl, L., Bessagnet, B., Favez, O., Garziandia, E.L., 2015. Episodes de pollution par-
Macintyre, H.L., Heaviside, C., Neal, L.S., Agnew, P., Thornes, J., Vardoulakis, S., 2016. ticulaire en France : quels enseignements tirer des re
cents e

pisodes ? Particu-
Mortality and emergency hospitalizations associated with atmospheric partic- late air pollution episodes in France : lessons learnt from recent episodes. Poll.
ulate matter episodes across the UK in spring 2014. Environ. Int. 97, 108e116. Atmos., Nume
ro spe
101e114. Retrieved from. http://www.lcsqa.org/system/
https://doi.org/10.1016/j.envint.2016.07.018. files/imagecache/ib_thumbnail/images/extrait_ineris-pollution_
O'Driscoll, R., ApSimon, H.M., Oxley, T., Molden, N., Stettler, M.E.J., Thiyagarajah, A., atmospherique_mars_2015.pdf.
2016. A Portable Emissions Measurement System (PEMS) study of NOx and € din, Å., Jerksjo
Sjo €, M., Fallgren, H., Salberg, H., Parsmo, R., Hult, C., Yahya, M.-R.,
primary NO2 emissions from Euro 6 diesel passenger cars and comparison with Wisell, T., Linde
n, J., 2017. On-Road Emission Performance of Late Model Diesel
COPERT emission factors. Atmos. Environ. 145, 81e91. https://doi.org/10.1016/j. and Gasoline Vehicles as Measured by Remote Sensing.
atmosenv.2016.09.021. TfL, 2011. Motorcycles to be allowed permanent access to bus lanes [WWW
Paris, M. de, 2017. Le bilan des de
placements en 2016 a  Paris. http://temis. Document]. URL. https://tfl.gov.uk/info-for/media/press-releases/2011/
documentation.developpement-durable.gouv.fr/docs/Temis/0033/Temis- december/motorcycles-to-be-allowed-permanent-access-to-bus-lanes.
0033289/8724_2016.pdf. Weiss, M., Bonnel, P., Hummel, R., Provenza, A., Manfredi, U., 2011b. On-road
Petit, J.E., Amodeo, T., Meleux, F., Bessagnet, B., Menut, L., Grenier, D., Favez, O., 2017. emissions of light-duty vehicles in Europe. Environ. Sci. Technol. 45,
Characterising an intense PM pollution episode in March 2015 in France from 8575e8581. https://doi.org/10.1021/es2008424.
multi-site approach and near real time data: Climatology, variabilities, Weiss, M., Bonnel, P., Kühlwein, J., Provenza, A., Lambrecht, U., Alessandrini, S.,
geographical origins and model evaluation. Atmos. Environ. 155 (March 2015), Carriero, M., Colombo, R., Forni, F., Lanappe, G., Le Lijour, P., Manfredi, U.,
68e84. https://doi.org/10.1016/j.atmosenv.2017.02.012. Montigny, F., Sculati, M., 2012. Will Euro 6 reduce the NOx emissions of new
Pham, C.T., Kameda, T., Toriba, A., Hayakawa, K., 2013. Polycyclic aromatic hydro- diesel cars? e Insights from on-road tests with Portable Emissions Measure-
carbons and nitropolycyclic aromatic hydrocarbons in particulates emitted by ment Systems (PEMS). Atmos. Environ. 62, 657e665. https://doi.org/10.1016/j.
motorcycles. Environ. Pollut. 183, 175e183. https://doi.org/10.1016/j.envpol. atmosenv.2012.08.056.
2013.01.003. Yang, H.-H., Chien, S.-M., Chao, M.-R., Lin, C.-C., 2005. Particle size distribution of
Platt, S.M., Haddad, I. El, Pieber, S.M., Zardini, A.A., Clairotte, M., Barmet, P., polycyclic aromatic hydrocarbons in motorcycle exhaust emissions. J. Hazard
Pfaffenberger, L., Wolf, R., Slowik, J.G., Fuller, S.J., Kalberer, M., Chirico, R., Mater. 125, 154e159. https://doi.org/10.1016/j.jhazmat.2005.05.019.
Dommen, J., Astorga, C., Zimmermann, R., Marchand, N., Hellebust, S., Zhang, S., Wu, Y., Hu, J., Huang, R., Zhou, Y., Bao, X., Fu, L., Hao, J., 2014. Can Euro V
Baltensperger, U., 2014. Two-stroke scooters are a dominant source of air heavy-duty diesel engines, diesel hybrid and alternative fuel technologies
pollution in many cities 1e7. https://doi.org/10.1038/ncomms4749. mitigate NOX emissions? New evidence from on-road tests of buses in China.
Robert, M.A., VanBergen, S., Kleeman, M.J., Jakober, C.A., 2007. Size and composition Appl. Energy 132, 118e126. https://doi.org/10.1016/j.apenergy.2014.07.008.
distributions of particulate matter emissions: Part 1 - Light-duty gasoline