In order to identify and quantify key species associated with non-exhaust
emissions and exhaust vehicular emissions, a large comprehensive dataset of
particulate species has been obtained thanks to simultaneous near-road and
urban background measurements coupled with detailed traffic counts and
chassis dynamometer measurements of exhaust emissions of a few in-use
vehicles well-represented in the French fleet. Elemental carbon, brake-wear
metals (Cu, Fe, Sb, Sn, Mn),
Traffic is a major source of particulate matter in urban environments through both exhaust and non-exhaust emissions. Thanks to stringent regulations, vehicles with more efficient catalytic converters and diesel particle filters have been progressively introduced into the European fleet. As a consequence particulate exhaust emissions have strongly decreased and non-exhaust particulate emissions (particles resuspended by moving traffic and those from the wear of brakes, tyres, road surface, etc.) will contribute to the major part of particulate vehicular emissions in the near future (Amato et al., 2014). Current research estimates that non-exhaust emissions already substantially contribute to traffic emissions, with differences between sites due to local meteorology, local emissions or traffic characteristics (e.g. Omstedt et al., 2005; Thorpe and Harrison, 2008; Bukowiecki et al., 2009; Lawrence et al., 2016). Even with electric vehicles, traffic will continue to be a source of particulate matter (PM) through non-exhaust emissions (Pant and Harrison, 2013; Timmers and Achten, 2016).
Also, knowledge on the deleterious impacts of PM from vehicular emissions on
human health is increasing. There is now strong evidence that
traffic-related PM is responsible for adverse health effects due to the
health effect of both carbonaceous material from exhaust emissions and
redox-active metals in traffic-generated dust including road, brake and tyre
wear (Kukutschová et al., 2009; Cassee et al., 2013; Amato et al., 2014,
and references therein; Pardo et al., 2015; Poprac et al., 2017). Recently,
on the one hand, Shirmohammadi et al. (2017) and Weber et al. (2018) have
shown the important role of non-tailpipe emissions in the oxidative
potential of particulate matter species identified as tracers of vehicle
abrasion; and, on the other hand, the health impacts of coarse particles is
now better documented (Beelen et al., 2014; Cheng et al., 2015; Malig et
al., 2013). Therefore, a better knowledge on vehicular emissions is required
to better understand their contribution to urban atmospheric PM
Chassis dynamometer measurements allow the determination of exhaust vehicular emissions under controlled testing conditions, but, because of high costs, these tests often include small sets of vehicles that cannot be representative of large variation in engine type, age and maintenance history. They also do not represent the variability in driving types in various environments. All of these are parameters that strongly influence real-world vehicular emissions. Additionally, these tests cannot simulate accurately the effect of dilution on particle equilibrium in the road atmosphere (Kim et al., 2016), the possible rapid-aging effects (e.g. Platt et al., 2017) and non-tailpipe emissions (e.g. Thorpe and Harrison, 2008). For all these reasons, it is thought that more realistic estimates of vehicle emissions are determined from air quality measurements in the near-road atmosphere (Phuleria et al., 2017), including car-chasing, tunnel or roadside measurements (Pant and Harrison, 2013, and references therein; Jezek et al., 2015).
In this study, a large comprehensive dataset on the chemical composition of
PM
Vehicles were operated on a chassis dynamometer, and exhaust emissions were measured using a sampling train including a constant volume sampling (CVS) system in which emissions are diluted with filtered air and sampled using quartz filters and polyurethane foams (PUF).
Five vehicles representative of the most frequent vehicle classes in the French fleet in circulation were selected (Euro 3 diesel, Euro 4 diesel, Euro 4 diesel retrofitted with a particle filter, Euro 2 petrol, Euro 4 petrol). They were in-use private cars since rental vehicles may not be representative of the national fleet (low mileage). All vehicles were operated using commercial diesel and petrol fuels. Selection criteria and characteristics of tested vehicles are described in the Supplement.
Driving cycles designed to be representative of real driving conditions in
Europe are performed in this study (André, 2004). The ARTEMIS urban cycle
represents driving conditions in urban areas (repeated acceleration and
vehicle speed below 40 km h
Vehicles were operated on a chassis dynamometer using a CVS system to dilute exhaust emissions with filtered ambient air. The filtration system included four filters and a cartridge in series: M6-F7-F9, M5, F7 EN-779-2012 filters; a HEPA H13 EN1822-2009 filter; and a cylindrical cartridge of charcoal scrubber.
Regulated emissions were determined with continuous monitors for CO,
Analysis of test blank filters (collected following the same procedure as vehicle test filters including driving cycle durations with filtered air) showed contaminations from the CVS system. It was obvious that some organic compounds measured in test blanks result from the desorption of semi-volatile organics deposited into the CVS, which is favoured by cleaner air. Since blank levels decreased in time with passing air dilution through the system, our procedure included such a step maintained for a duration corresponding to at least two driving cycles before each vehicle test. Test blanks are used to correct measurements.
The joint PM-DRIVE (Particulate Matter Direct and Indirect on-road Vehicular Emissions) and
MOCOPO (Measuring and mOdelling traffic COngestion and POllution) field
campaign took place from 9 to 23 September, 2011. It included
meteorology and traffic measurements, near-road and urban background PM
The Grenoble conurbation is a large city with about 700 000 inhabitants. It
is located in the southeast of France in the French Alps and is surrounded
by three mountain ranges (Vercors, Chartreuse, and Belledonne). The traffic
site was located (45.150641
On both traffic and urban background sites, PM
Traffic counters (double electromagnetic loops) were installed in order to identify the passing of all vehicles, the length of their chassis and their speeds, the determination of the two vehicle classes used in this study (light-duty and heavy-duty vehicles, or LDVs and HDVs), and the identification of periods of stop-and-go or flowing traffic. Vehicle mean speeds are computed as the harmonic mean speed of the cars passing over the detector (flow speed, i.e. the spatial mean speed) (Hall, 2001).
Traffic cameras mounted on a roadway gantry were also used to monitor
traffic at the measurement site. They were used to capture the license plate
numbers of passing vehicles. Plate numbers were later used to classify
vehicular traffic into different categories: Euro standards and fuel type
(diesel or petrol). The traffic of the highway is detailed elsewhere (DeWitt
et al., 2015; Fallah Shorshani et al., 2015). On weekdays, the average
hourly traffic included 2850 diesel vehicles (including about 200
heavy-duty vehicles) and 1025 petrol vehicles. The harmonic vehicle mean
speed was about 80 km h
The vehicle fleet was close to the national one for the year 2011, with 72 % diesel vehicles and Euro 3 and 4 vehicles representing most of the vehicles (30 % and 36 %, respectively). Virtually all heavy-duty vehicles are diesel.
A Young meteorological station was installed at the traffic site to capture wind speed and direction, while relative humidity, temperature data and rain data are obtained from two stations located in the Grenoble conurbation (see Supplement Sect. IV).
The wind speed was low during the field campaign (on average 0.98 m s
The measurements of carbonaceous material (EC and OC) in PM samples were performed using the thermo-optical transmission (TOT) method on a Sunset Lab analyser (Jaffrezo et al., 2005; Aymoz et al., 2006) following the EUSAAR2 temperature protocol (Cavalli et al., 2010). Ionic species were analysed with ionic chromatography (IC) following a well-established method (Jaffrezo et al., 1998; Waked et al., 2014). Metals were analysed using inductively coupled plasma mass spectrometry (ICP-MS) (Waked et al., 2014). The chemical speciation of organic particles is performed by gas chromatography–mass spectrometry (GC-MS), except for PAHs that were measured by liquid chromatography (HPLC) using a fluorescence detector (Piot, 2011; Golly et al., 2015).
Hence, a large number of particulate chemical species have been measured in
the filter samples simultaneously collected at the traffic and urban
background sites, including EC, OC, 9 major ions (
The same array of chemical species were also quantified in the filter samples from the chassis dynamometer experiments.
A series of multivariate data analysis tools have been used in order to
define which species are related to traffic, to identify influential
parameters and to quantify their respective influences. Thanks to
simultaneous measurements at near-traffic and background sites, local
increments in concentration at the traffic site have been calculated as the
difference between near-traffic and urban background concentrations. The sign
and Wilcoxon signed-rank tests have been used to estimate if concentrations
measured at the near-traffic site are significantly higher than
concentrations measured at the urban background site and can possibly be
ascribed to local traffic emissions. As a complementary indication of
the relation to traffic, Spearman correlations with traffic data (total traffic,
light-duty traffic, heavy-duty traffic), with
Continued.
Increments in concentrations for all species strongly associated with
traffic are transformed into emissions according to a well-established
procedure (e.g. Pant and Harrison, 2013; Charron and Harrison, 2005; details
in the Supplement Sect. V). This method enables the calculation of average
emission factors for the mixed traffic fleet of the RN87 highway assuming
that (1) increments in concentration (near-traffic site minus urban traffic
site) are from local traffic, (2) emissions of
Average concentrations measured at the traffic site are presented in the
Supplement (Sect. II). Four-hour PM
Median concentrations measured at the roadside site (Echirolles)
and urban background site (Les Frênes) as well as comparison with respective
median TEOM-FDMS PM
Table 1a and b present species for which concentrations are significantly
higher at the near-traffic site and assumed from local emissions, and
Fig. 2 presents median concentrations measured at both sites. These species could
be distributed into two main groups according to both the significance of
the contribution of traffic to atmospheric levels and the strength of
relations with traffic indicators (traffic counts,
Cu, Fe, Mn, Sb and Sn have concentrations strongly correlated with traffic
indicators (Table 1a). Their local increments in concentration due to
traffic range from 54 % to 84 %. The concentrations of these species are
linearly related to each other with near-zero intercepts (Fig. 4),
confirming that they come from the same source. Similarly to this study,
Amato et al. (2011a) and Harrison et al. (2012) measured strong increments
for Fe, Cu, Sb and Sn concentrations at traffic sites with strong
correlations between them. Here, Cu, Fe and Sn are the metals that are the
most closely related (Pearson
The 4 h concentrations measured at the traffic site:
Scatterplots and linear relationships for a few species:
Cr and Ti are also species significantly correlated to traffic indicators but to a lower extent than Cu, Fe, Mn, Sb and Sn. Cr (Boogaard et al., 2011; Amato et al., 2011a) and Ti (Amato et al., 2011a) also showed higher atmospheric concentrations at street locations than at urban background sites in previous studies. Cr and Ti show a behaviour different from that of other elements coming from brake wear. They present sharp peaks in the morning, poorer correlations with copper and significant correlations with Al (Fig. 4 and Table 1a). When the very high morning concentrations are removed (mornings of workdays), their temporal variations are much closer to the ones of metals from brake-wear emissions. This suggests that another source influenced Cr and Ti concentrations near the traffic site in the morning, possibly nearby metalworking activities. Similarly, while morning peaks are less obvious in the temporal variations, the exclusion of morning data for Fe and Mn improves their correlations with Cu (Fig. 4).
Many of these species are metals that are known to arise from brake-wear emissions (Thorpe and Harrison, 2008; Pant and Harrison, 2013; Grigoratos and Martini, 2015). Indeed, Fe could come from the lining (steel or iron powder) in semi-metallic brakes, from fibres such as steel, or cast iron rotor wear for slightly metallic brakes. Cu is a high-temperature lubricant present in linings and it is also included in fibres as brass to increase braking performance. Sb is an element of brake lining both in filler as antimony sulfate and in lubricant as antimony trisulfide. Chromium oxides are elements of the filler of brake linings used for their thermal properties, and potassium titanate fibers are present as a strengthener in organic linings (Sanders et al., 2003; Grigoratos and Martini, 2014, and references therein). Cr could also come from lubricant oil combustion (Pulles et al., 2012).
Only three light-molecular-weight PAHs (An, Fla, Pyr) are strongly associated with traffic indicators even though only the particulate phase is determined (Table 1b). It is well-established that light-molecular-weight PAHs are emitted by diesel vehicles, while higher-molecular-weight PAHs are rather associated with petrol vehicle emissions (Zielinska et al., 2004; Phuleria et al., 2006, 2007; Pant and Harrison, 2013). Not surprisingly, in this site dominated by diesel vehicles, high-molecular-weight PAHs (from five rings) are more significantly correlated with levoglucosan, suggesting a closer relation to biomass burning emissions than to on-road petrol vehicles. Proportional concentrations of An, Pyr and Fla suggest that they mainly come from the same source, likely diesel exhaust emissions (Figs. 3 and 4). Specifically, Pyr and Fla show high increments in concentration (as an indication of the low contribution of the urban background compared to strong traffic contribution, and/or rapid photochemical degradation) and a strong linear relationship without any significant intercept (as an indication of common origin). Concentrations of An have more variability and larger contribution from the urban background.
The
Results of the multiple linear regressions with the heavy-duty
traffic (HDV) and the light-duty traffic (LDV): square correlation
coefficients, unstandardized coefficients with standard deviations for HDV
and LDV,
This second group corresponds to species with local increments below 50 %
and no significant, or only weak, correlations with traffic indicators. It
is highly heterogeneous since it includes OC,
Not surprisingly
OC concentrations are significantly higher at the traffic site (
Ba and Co concentrations are significantly higher at the traffic site
(
The concentrations of Phe, BaA, C18 and C27–C33 alkanes are significantly
higher at the traffic site and some of them are significantly correlated
with
Average emissions factors (median and IQR: interquartile range)
determined by chassis dynamometer measurements for different types of
passenger cars: Euro 3 diesel (E3D), Euro 4 diesel (E4D), Euro 2 petrol
(E2P), Euro 4 petrol (E4P) and Euro 4 diesel equipped with a particle filter
(E4D
Average emission factors (EFs) are presented in Table 2a, b, c and d. Most EFs
show large standard deviations. This variability reflects the presence of
vehicles with various emission levels (diesel/petrol; different standards
and engine load; cold start/hot vehicles; presence of a few high-emitting
vehicles). It could also be related to the variability in the vehicle fleet
(on average 5 % heavy vehicles but ranging from 0.3 % to 12 %) and to
the various traffic conditions (from fluid with speeds up to 90 km h
Table 3 presents average EFs for heavy-duty and light-duty traffic, their standard deviations and confidence intervals at 95 %.
The traffic-fleet EF for EC determined in this study (39 mg veh
The average light-duty-traffic-fleet EF for EC is in excellent agreement with the EFs of Euro 3 and Euro 4 diesel vehicles obtained from chassis dynamometer measurements in our study, for which the highest EFs were clearly observed when vehicles are cold (Fig. 5a). Indeed, these two types of vehicles represented the largest proportion of passenger cars on the RN87 freeway in 2011 (Fallah Shorshani et al., 2015), as well as in the French national fleet. Similarly to the results for diesel vehicles tested by Fujita et al. (2007) and Lough et al. (2007), EC has the highest emission factor in diesel exhaust. It also could be noted that the heavy-duty-traffic-fleet EF for EC is about 5 times higher than the one determined for light-duty traffic.
Emission factors for EC
The emissions of OC could not be discriminated between light-duty and heavy-duty traffic (coefficients not significantly different from zero). However, it can be observed that the traffic-fleet EF for OC is larger than could be expected from exhaust measurement of test vehicles (Table 4, Fig. 5b). Note that traffic-fleet EFs for OC from other studies (Handler et al., 2008; Alves et al., 2015; Cui et al., 2016; He et al., 2008) are at least as high as ours, and the EFs for exhaust OC from Cheung et al. (2010) for Euro 4 diesel vehicles (with and without a diesel particulate filter) are similar to ours. There are many likely explanations for the traffic-fleet EF for OC being higher than expected: proportionally larger contribution of the heavy-duty traffic to OC than to the EC, contribution of non-exhaust emissions to the OC (e.g. tyre wear), contribution of high-emitting vehicles, and rapid formation of secondary OC in the roadside atmosphere. Further studies are required to assess the respective importance of these processes. In particular, a better knowledge of the particle size distribution of OC emitted by traffic might be useful.
Average
Ratios for roadside, incremental (roadside minus urban background) concentrations, and RN87-traffic emission factors, with standard deviations.
Fe presents by far the third highest traffic emission rate after those of EC
and OC (6.7 mg veh
Our results can be compared with other European traffic EFs determined for
PM
EFs for brake-wear metals (Cu, Fe, Sb and Sn) are 8 to 13 times higher for the heavy-duty traffic than for the light-duty traffic (Table 3). The estimations from Bukowiecki et al. (2009) for brake wear only are much lower than ours but somewhat proportional (factors of 4.6 to 6.8 for light-duty EFs and from 4.3 to 8.5 for the heavy-duty traffic – Sb excluded). This consistency between brake profiles suggests that brake compositions would be similar in different European countries. Traffic-fleet light-duty EFs for brake-wear metals are much higher than the ones from chassis dynamometer exhaust measurements (this study: Table 4; Cheung et al., 2010), confirming the dominant contribution of braking for these elements.
The sum of traffic-fleet EFs for metals related to brake wear (Ba, Cr, Cu,
Fe, Mn, Sb, Sn, Ti – Table 2b) leads to a total of 7.3 mg km
According to inventories atmospheric copper is largely from brake wear.
Indeed brake wear represents 50 %–75 % of European Cu emissions (Denier Van
der Gon et al., 2007) and 64 % of French Cu emissions (CITEPA, 2018).
Since Sb is another well-known constituent of brake with very few other
atmospheric sources, the
The very strong linear relationships found between Fe, Cu, Mn and Sn with no
significant intercept (virtually equal to zero) suggest that ratios
including these compounds are also worthy of attention. Even though Fe is much less
specific of brake-wear emissions than Cu or Sb, Hulskotte et al. (2014)
observed a stable
Ratios with Mn and Sn are more rarely discussed in the literature, and
published information is scarce.
As previously observed (Schauer et al., 2002; Perrone et al., 2014; El
Haddad et al., 2009),
In agreement with other chassis dynamometer measurements (Rogge et al.,
1993a; Cheung et al., 2010; Perrone et al., 2014; Cui et al., 2017), the
most important
The relative contributions of the
Low-molecular-weight PAHs (An, Fla, Pyr) are frequently associated with
particulate diesel exhaust emissions (e.g. Kleeman et al., 2008; Keyte et
al., 2016). Accordingly, both chassis dynamometer and in situ measurements indicate
that diesel vehicles are a major source of these chemical species. Indeed,
they are almost absent in petrol exhaust emissions and are strongly related
to RN87 traffic emissions. EFs for PAHs determined for the heavy-duty
traffic are much higher than the ones determined for light-duty traffic:
from 6 times higher for Fla up to 20 times higher for Pyr (Table 3). High
emissions of Pyr by heavy-duty vehicles have already been observed anywhere
else (Liacos et al., 2012; Cui et al., 2017). While the comparison between
EFs determined during different conditions is not straightforward in the absence
of gas-phase measurements due to the high vapour pressure of the
lowest-molecular-weight organics (Fujitani et al., 2012; Polo-Rehn, 2013), quite
good agreements are found between chassis dynamometer measurements and
traffic EFs and with data from other studies. Indeed, again the light-duty-traffic EFs for Fla and Pyr are quite consistent with the ones of the diesel
Euro 3 measured with the chassis dynamometer. The light-duty-traffic EF for An
is proportionally lower, somewhere between the ones of test Euro 3 and Euro
4 vehicles. The average EF for Pyr for the Euro 3 diesel vehicle of this
study (Table 4) is close to the ones of Perrone et al. (2014) for Euro 3
diesel vehicles (
In agreement with previous observations (Phuleria et al., 2006, 2007; He
et al., 2008; El Haddad et al., 2009; Alves et al., 2016; Pant et al.,
2017), 17
The average traffic-fleet EFs for
Since EU tyres contain about 1 % zinc oxide (Pant and Harrison, 2013) and Zn is the most abundant metallic element in tyres commercialized in the US (Apeagyei et al., 2011), Zn is often proposed as a key tracer of tyre wear emissions. In this study, similarly to other works (Boogaard et al., 2011; Amato et al., 2011a), Zn concentrations did not show any roadside increment. This suggests the prevalence of other sources of Zn in the Grenoble Alpes conurbation, as well as in other urban environments. Further research is needed to determine proper tracers of tyre wear emissions.
Thanks to a very large comprehensive dataset of particulate species collected from a simultaneous near-road and urban background measurement field campaign and chassis dynamometer experiments of a few in-use passenger cars, this study was able to determine emission factors for many particulate species from road traffic and to identify and quantify tracers of exhaust and non-exhaust vehicular emissions that could be used in source apportionment studies. Near-road measurements are made near a freeway with various driving conditions from free-flowing to stop-and-go traffic, including frequent and severe braking events during periods of congestion (morning and afternoon commuting times of workdays).
EC has the highest traffic emission factors and is strongly associated with
diesel traffic. The emission factor for EC for the light-duty traffic is
similar to the ones of passenger diesel cars without particle filters. EC
emissions from heavy-duty vehicles are estimated to be 5 times higher than
those for light-duty vehicles. The traffic-fleet EF for OC is slightly
larger than those deduced from exhaust measurement of test vehicles. This
later observation would require further investigations in order to delineate
the several possible causes for such observation. The determination of the
particle size distribution of OC could improve knowledge of the organic
emissions of traffic. In this environment dominated by the diesel traffic,
the
Results showed the important contribution of metals from brake wear to
particulate vehicular emissions. In particular, Fe has the third highest
traffic emission factor after EC and OC. Total brake-wear emissions are
estimated for the RN87 highway: they are on average almost twice the
particle emission standards for the exhausts of newer vehicles (from Euro
5). We have shown that Cu is another important contributor to PM
Particulate organic emission data for European motor vehicles are scarce. In
this study, a few PAHs,
Hopanes are markers of lubricating oil in the emissions of high-emitting
vehicles (Rogge et al., 1993a; Zielinska et al., 2004). In this study two
hopanes (17
This study determines many quantitative data of traffic exhaust and non-exhaust emissions that could help in a better definition of traffic emissions in source apportionment studies.
Data are available upon request.
The supplement related to this article is available online at:
AC was responsible for the design and coordination of the PM-DRIVE project and the data analyses and designed the vehicle experiments. JLJ, JLB and NM contributed to the design of the PM-DRIVE project. CB was responsible for the design and coordination of the MOCOPO project and designed the traffic measurements. JLJ was responsible for the field experiments and PM analyses (metals, ions, EC, OC). JLB was responsible for organic speciation. HC and GG were responsible for regulated PM and gas sampling. LPR and BG contributed to field measurements, PM analyses and interpretation of data. The paper was prepared by AC. JLJ, JLB, BG, LPR, NM and CB contributed to the interpretation of the results and discussions on the paper.
The authors declare that they have no conflict of interest.
This work was funded by CORTEA-ADEME (PM-DRIVE program 1162C0002), which includes the funding of chassis dynamometer and near-road field and urban background campaigns, and PREDIT (MOCOPO program), which includes the near-road field measurements of regulated pollutants and traffic characteristics. Lucie Polo-Rehn's PhD was funded by the Région Rhône-Alpes. Rain data were supplied by Météo France. The authors would like to thank Patrick Tassel, Pascal Perret and Mathieu Goriaux (chassis dynamometer experiments) as well as Julie Cozic and Jean-Charles Francony (sample analyses) for their contribution to this work. We also thank Michel André for supplying COPCETE emission factors. Part of the chemical analysis was performed on equipment provided by Labex OSU@2020 (ANR10 LABX56). Edited by: Sally E. Pusede Reviewed by: two anonymous referees