Introduction
Traffic is a diverse and important source of air pollution and is complex to
describe in terms of per vehicle emissions. The amount of emitted pollutants
depends on individual vehicle parameters, the engine type and displacement,
the type of exhaust after-treatment system, fuel quality, maintenance
status, traffic situations, topography, driver behavior and weather
conditions. Owing to the large number of variables, different statistical
analyses and measurement approaches have been employed in order to evaluate
traffic emissions. These vary in complexity in terms of describing traffic
activity and emission factor (EF) determination. Franco et al. (2013) define
EFs as being different empirical functional relations of emitted pollutants to the
activity that causes them. Most standardized and robust EFs were found to be
produced in laboratories using dynamometer tests with prescribed driving
cycles. These tests can produce the following: (a) aggregated or bag results with respect
to the mean speed or some other kinematic parameter (e.g. mean acceleration)
of a driving cycle; or (b) instantaneous emission data, where the emissions
values measured can be related to recorded instantaneous kinematic or engine
covariates (Perrone et al., 2014). But the nature and conditions of the
tests limits both the number of vehicles tested and the application to many
on-road or so-called “real-world” conditions. In order to validate the
emission model predictions and to compare their performance to actual
vehicle emissions, “real-world” EF measurement techniques have been
developed (Franco et al., 2013). These employ different techniques for
measuring numerous vehicles in use in actual traffic situations: the
measurements were performed through the use of remote sensing next to the
roads, following vehicles on the roads, the use of on-board diagnostics
data, or from data taken in tunnels (some of the first such experiments may
be found in Bishop et al., 1996; Hansen and Rosen, 1990; Weingartner et al.,
1997).
The various “real-world” methods have been described as being less precise
than the dynamometer studies because the tests are not as repeatable as
their dynamometer counterparts owing to the absence of standard cycles and
the presence of additional uncontrolled parameters introducing variability,
such as environmental or traffic conditions, driver behavior or highly
transient operations (Franco et al., 2013). The on-road measurements have
some inherent drawbacks. Two possible shortcomings are that the
remote-sensing method can provide only a snapshot of the vehicle emissions
and not how the emissions vary during the trip (Franco et al., 2013) and
that the on-road chasing method cannot be used in dense traffic
situations, where emissions from other vehicles would disturb the
background measurements (Ježek et al., 2015; Wang et al., 2011). Their
advantage over laboratory measurements is that, over a short period of time,
a large number of in-use vehicles can be measured and a representative
emission factor distribution for different vehicle categories can be
obtained. Most of the previous on-road BC emission factor measurements for
individual vehicles were performed on diesel-fueled trucks and on cars with
the spark ignition engine, henceforth referred to as gasoline cars
(Ban-Weiss et al., 2009; Dallmann et al., 2011, 2012, 2014; Hansen and
Rosen, 1990; Wang et al., 2011, 2012). Many of these studies revealed that a
small percentage of vehicles – the so-called super emitters contribute
disproportionately to total vehicle emissions. Ban-Weiss et al. (2009)
demonstrated that 10 % of the trucks contributed 40 % of the BC and PN
emissions. Wang et al. (2011) showed that, in their measured fleet, 20 %
of the trucks contributed 50 % of the carbon monoxide (CO) and PN0.5
emissions, 60 % of the PM0.5 (the particle number concentration –
PN; and particulate mass concentration (PM) subscripts denote here the
largest mobility diameter [µm] of aerosol particles measured, in this
case aerosol particles of 0.5 µm and smaller) and over 70 % of
black carbon (BC) emissions. Bishop and Stedman (2008) report the same trend
for nitrogen oxides (NOx), CO and hydrocarbons (HC). The advantage of
individual vehicle measurements over average fleet emission factors, as is
often expressed by dynamometer or portable emission measurement system
(PEMS) studies, is the ability to detect and express the distribution of
emissions from many vehicles as well as to identify “super emitters” and
their contributions within the vehicle population, serving as a basis for
the implementation of improved emission data, more efficient abatement
strategies and monitoring of progress on controls.
The chasing method allows us to capture a range of EF from a single vehicle
and to measure the EF distribution rather than just a single value as is
recorded with the stationary method. Depending on engine operation state,
each vehicle produces a range of EF with most values around a representative
value (median) and a long super emission tail – the comparison of the
chasing method and the stationary method can be found in Ježek et al. (2015). With a single stationary measurement we can capture only a single
value of the vehicle's EF distribution and several repetitions of a vehicle
would be necessary to obtain that vehicle's EF distribution. We believe that
using a single vehicle's EF distribution measured in real driving conditions
and using the collective distribution of the vehicle fleet to model traffic
emissions could improve model predictions. Knowing the EF representative
value and the super emission tails allows quantifying the effect of
potential abatement measures, e.g. how changing a driving regime would
influence emissions at a certain section of the city. Previous studies using
the chasing method for EF measurements in real driving conditions were
performed on fleets of buses, light duty vehicles (LGV) with gasoline
engines and heavy-goods vehicles (HGV) with diesel engines (Canagaratna et
al., 2004; Herndon et al., 2005; Schneider et al., 2008; Shorter et al.,
2005; Wang et al., 2011, 2012). Shorter et al. (2005) discuss the
effectiveness of the NOx emission reduction in different engine and
exhaust system technologies, which had been introduced to the New York bus
fleet. They found that NOx emissions from diesel and Compressed Natural
Gas (CNG) buses were comparable and that diesel hybrid electric buses had
approximately one-half the NOx emissions. They also found that in the
group of diesel buses equipped with continuously regenerating technology
(CRT), nitrogen dioxide (NO2) represented a third of emitted NOx,
while in non-CRT buses emissions, the percentage of NO2 was less than
10 %. Similar NO2 to NOx ratios were found by Carslaw and
Rhys-Tyler (2013), who used a remote-sensing technique to measure the
emissions of almost 70 000 vehicles in the United Kingdom (UK), where 30 % of
NOx were emitted as NO2 by Enhanced Environmentally friendly
Vehicles (EEV). The EEV is a recommended standard in the European Union for
HGVs with lower PM emission values than a Euro VI vehicle but the same
NOx standard as a Euro V.
Wang et al. (2011) measured the EF of BC, CO and PM0.5 on a fleet of
230 trucks and 57 buses in China, and identified “heavy emitters” in the
road fleet. They found that 5 % of the trucks contributed 50 % of the BC
emissions, and 20 % of the trucks contributed 50 % of the CO and
PM0.5 emissions. Furthermore they found that the EFs of trucks
registered outside Beijing were significantly higher than those that were
subject to the stricter engine and fuel quality standards enforced in
Beijing. Because numerous trucks registered outside Beijing operate in the
Beijing area, restricting Beijing-registered truck emissions is not
sufficient to reduce traffic-related pollution in the city. Their bus fleet
measurements showed that replacement of older buses with newer buses (Euro IV
and CNG) compared to their predecessors (Euro II and Euro I) were indeed
an effective way to reduce the emissions of the measured pollutants. In
their follow-up study (Wang et al., 2012), they employed the same method on
a fleet of 440 on-road trucks, measuring the EF of NOx and BC. They
found that the measures taken in Beijing were effective for the BC emissions
of trucks that were from that area, but they did not observe such a trend
for NOx emissions.
An extensive on-road measurement study was performed in the UK by Carslaw
and Rhys-Tyler (2013). They employed a remote-sensing technique to measure
the emissions of NO, NO2 and NH3 on a fleet of almost 70 000
individual vehicles which included also vans, passenger cars with a
compression ignition engine (henceforth referred to as diesel cars), and
gasoline cars. Matching these to vehicle registration data, they found that
only gasoline-fueled vehicles had shown an appreciable reduction in NOx
emissions over the past 15–20 years, whereas diesel-fueled vehicles have
not. They found that there was an influence of vehicle manufacturer for Euro 4/5
vehicles and that Euro 4/5 diesel vehicles with smaller displacements
emit less NO than those with larger displacements. According to the European
Automobile Manufacturers' Association (ACEA) the motorization in Europe is
increasing for passenger cars and the commercial vehicle fleet – by about
50 % in 2 decades (1990–2010). Fleet trends show that the percentage
of diesel cars is also rising from about 30 % in 2000 to about 60 % in
2011, and that the most popular passenger cars by segment are small-sized and the lower segment of medium-sized cars which respectively represent 34.2 and 22.1 % of all new
cars sold in Europe in 2011 (ACEA, 2012). A slightly smaller percentage of
diesel cars (55 %) was reported by the European Environment Agency (EEA,
2013a) who, in their report titled “Monitoring CO2 emissions from new
passenger cars in the EU”. They state that the average car weight was at
its highest in the last 9 years, the average engine capacity had
decreased by 5 % since 2007, and, despite of these changes, the improved
vehicle technology has led to greater fuel efficiency and lower average
CO2 emissions per kilometer traveled (EEA, 2013b). This report was
based on data provided by the manufactures who were obliged to measure
CO2 emissions using the type approved test cycle (the New European Driving Cycle; NEDC) in laboratory
conditions. The statement was refuted by International Council on Clean
Transportation in their 2013 white paper (Mock et al., 2013); in which they
compared official and “real-world” fuel consumption and CO2 values for
cars in Europe and the United States. The report contained an assessment of
the results of several on-road driving data sets from various European
countries, where they found underestimation by type-approved measurements
relative to on-road CO2 emissions. Namely, in 2001, the discrepancy
between the two had been below 10 % and increased to around 25 % by
2011, with “real-world” emissions being higher than emissions according to
type-approval. The same report also clarifies that their analysis does not
suggest that manufacturers have done anything illegal. Instead it is
suggested that the NEDC was not appropriate to use for indicating fuel
consumption as it was originally not designed to measure this, nor was it
designed to measure CO2 emissions. Some features of the test procedure
can be exploited to influence test results, resulting in unrealistically low
CO2 emission levels. These issues are being addressed by the United
Nations through the development of a new vehicle test procedure, among other
things (Mock et al., 2013). Based on the limited availability of the data
that were used in previous studies, we postulate that using on-road emission
factors from a representative vehicle fleet could contribute significantly
to models' emission predictions. EF determination of a representative
on-road sample would include additional sources of variability which can be
controlled in the laboratory but not in real-world driving conditions.
BC, NOx and PN are emitted from internal combustion engines and
negatively impact people's health. The three pollutants do not have the same
formation process (Heywood, 1988; Kittelson et al., 2006). A more detailed
description may be found in Sect. S1 in the Supplement. It has been shown
that increased BC concentrations are a better indicator of hazardous health
effects of aerosol particle air pollution than the increase in the
legislated particle mass concentrations (Janssen et al., 2012); and that it
is after CO2 the second most important contributor to global warming
(Bond et al., 2013).
The research presented here is aimed to measure real-world BC EF of diesel
cars, since there was no previous research reporting BC EF of numerous
diesel cars measured individually in real driving conditions. Gasoline cars
and goods vehicles were included for comparison purposes. We also measured
vehicles' NOx and PN EFs due to their hazardous effects on health and
environment and for the comparison purposes to previous studies. We used the
chasing technique (Wang et al., 2011) and the running integration approach
to calculate individual vehicles EF (Ježek et al., 2015), because it
enables us to measure not only EFs of numerous individual in-use vehicles,
but also how their EFs change in time, giving us individual vehicle's EF
distribution. We analyze EF distribution within the vehicle category by
using the median EF value of individual vehicle's EF distribution and
compare our results to those of other chasing and remote-sensing studies. We
obtained registration information of the chased vehicles to demonstrate the
effects of vehicle age, vehicle maximum engine power, the ratio of maximum
power to vehicle size, and finally, the contribution of high emitters to the
total emissions of our measured fleet. We report the first on-road
determination of BC, NOx and PN EFs of passenger cars measured with the
chasing method and the first BC EFs of individual diesel cars measured in
real driving conditions.
Methodology
We performed our measurements in December 2011 over the course of 7 days on
Slovenian highways and regional roads, measuring predominantly the Slovenian
vehicle fleet (photographs from the measurement campaign are presented in
Supplement Fig. S1). Slovenia is a country positioned south of
the Alps, next to the Adriatic and opening to the Balkan and East European
region. Slovenian highways are part of the V. (Venice-Trieste/Koper-Ljubljana-Budapest-Kiev) and X. (Salzburg-Ljubljana-Zagreb-Belgrade-Thessaloniki) trans-European corridors
and are thus an important connection between central and east European
states, especially for the transport of goods. As a result, foreign vehicles
were also encountered and measured in our campaign.
In EF analysis we included any vehicle which emissions and background
concentrations we could capture without interference of other on-road
vehicles (vehicles that would drive in front of the chased vehicle). The
inclusion of the measurement in further analysis was determined on-road and
confirmed with video recordings of each chase. For most vehicles we measured
the background concentrations before and after the chase, in few instances
we used only one – measured before or after the chase. On average each chase
lasted for 2 and a half minutes, with the shortest chase lasting for
47 s
and the longest for 396 s. The traveling speed was changing within each
chasing episode but for most trucks it was between 80 and 90 kmh-1 and for
cars it was between 100 and 130 kmh-1. In the final analysis we excluded 10
cars because we could not obtain registration information needed to
categorize them as a diesel or a gasoline car.
Measurement instruments, their time resolutions, sampling flows and
measurement uncertainties.
Instrumentation
Species measured
Time resolution
Instrument flow
Measurement uncertainty
Carbocap GMP343 (Vaisala)
CO2
2 s
7 Lmin-1
3 ppm
Aethalometer AE33 (Aerosol d.o.o.)
BC
1 s
7 Lmin-1
30 ngm-3
FMPSa (TSI)
PN
1 s
10 Lmin-1
±10 to 20 %b
Nitric Oxide Monitor and an NO2 converter (models 410 and 401 of 2B Technologies)
NOx
10 s
0.7 Lmin-1
1.5 ppb
a Particle size range 5.6–560 nmb The uncertainty of PN measurements is calculated for each particle size
stage and varies within different stages. It is dependent on the measurement conditions and PN concentrations.
The mobile measurement platform used for the on-road chasing measurements is
described in detail in Ježek et al. (2015). We used instruments with high
time resolution (1 to 10 s) the Carbocap GMP343 (Vaisala) to measure
CO2, the Aethalometer AE33 (Aerosol d.o.o.),
the Fast Mobility Particle Sizer (TSI), for the on-road campaign we added
also a Nitric Oxide Monitor and an NO2 converter (models 410 and 401,
2B Technologies). For the Nitric Oxide Monitor the sampling line was a
Teflon tube, while for the rest we used static-dissipative tubing. The
instrumental details and measurement uncertainties are summarized in
Table 1. The Aethalometer data were compensated for
the loading effect using the Drinovec et al. (2015) compensation algorithm.
While the size distribution of the exhaust particles change with the engine
operation (Ježek et al., 2015; Sharma et al., 2005), a fact that might
have implications in the context of the health effects of exhaust particles,
Rayleigh–Debye–Gans theory (Sorensen, 2001; an example of such calculation
can be found in Kim et al., 2015) predicts the mass absorption cross-section
independent of the size distribution of the fractal aggregates. This is
consistent with the near-road observations by Ning et al. (2013).
Comparison of two integration approaches to calculate individual vehicle's
emission factor (EF). With the bulk integration, the EF is calculated by
integrating the plume from the beginning to the end of the chase; the median
EF is calculated with the running integration approach with 10 s integration
windows, from the EF distribution the median value is then calculated.
Emission factor calculation
We calculated the emission factor as the pollutant (P) per kg of fuel
consumed, assuming the equal dilution of all emitted pollutants and complete
combustion of the fuel, where almost all of the carbon in the fuel is oxidized
to CO2 (Ban-Weiss et al., 2009; Dallmann et al., 2011; Hansen and
Rosen, 1990), the fuel consumption can be estimated by measuring the
CO2 emissions.
EFP=∫tjtiPtj-Ptidta⋅∫tjtiCO2tj-CO2tidt⋅wc
The coefficient a in denominator represents the mass ratio between C and
CO2: a=12:44=0.2727, thus converting the mass concentration of
CO2 in Eq. (1) to units of mass concentration of C (mgCperm-3).
The carbon fraction in fuel wc for both gasoline and diesel was set to
0.86 (Huss et al., 2013). The subscripts ti and tj denote the time of the
beginning and end of integration step, respectively. NOx was treated as
NO2 equivalent with molar mass 46 gmol-1 (USEPA, 2010; Wang et
al., 2012). We used the running integration approach with the 10 s
integration step, to obtain individual vehicle's time-dependent EF, and thus
its EF distribution. From the distribution we calculated the median value
and used it as the representative EF value for the investigated vehicle.
The running integration approach is described in more detail in Ježek et
al. (2015), where the chasing method has been tested on contemporary cars in
controlled conditions. The results of the two integration approaches – the
bulk integration from the beginning and to the end of the chase (Wang et
al., 2011) and running integration, have already been compared in Ježek
et al. (2015). Here we again perform the comparison on a larger number of
measured vehicles. The regressions between the two methods for all three
investigated pollutants (BC, NOx and PN) are presented in
Fig. 1. For all three pollutants the Pearson's r′
coefficient was at least 0.97, all three intercepts were close to zero. The
bulk integration method gives somewhat larger EFs than the running
integration for BC and PN, while the slope is very close to unity for
NOx. Whilst BC and PN bulk integration overestimated the median EF by
9 and 14 % respectively, the bulk integration for NOx EF
underestimated the median EFs by 2 %. The slight underestimation of bulk
NOx EFs was probably because the instrument for NOx measurements
had lower time resolution (10 s) than other instruments (1 or 2 s), thus
super emission peaks were not as clearly resolved as they were for BC and
PN.
The uncertainty of the median value, which we here use as the representative
EF value for a single vehicle, was estimated to be -24/+26 % (Ježek
et al., 2015). This uncertainty is reduced when calculating the fleet EF
distribution. The uncertainty of the single measurement depends on the
measured CO2 and its signal to noise ratio (Ježek et al., 2015). We
constrain the calculation of the time evolving EF when CO2 values are
low by using a 10 s integrating time interval instead of shorter intervals.
This smooths out the high engine emission peaks, which are already smoothed
out by traveling through the exhaust system and the atmosphere to the
measurement instruments (Ajtay et al., 2005), and constrains the calculation
error, yet keeping the calculated median value unchanged. The dilution does
not affect the measurements of the single vehicle EF as long as the
CO2
increase is above the limit reported in Ježek et al. (2015). We show
this in a comparison between a PEMS measurement and a chasing determination
of EF (Fig. S2, data from Ježek et al., 2015). The impact of limited
number of vehicles was investigated by Ban-Weiss et al. (2009), where they
show that sampling ≥30 trucks should be a large enough sample.
Vehicle classification and fleet description
We collected license plate numbers and gained more information on the
measured vehicles from their registration certificates. The data provided by
the Slovenian Ministry of Infrastructure and Spatial Planning contained
information about each vehicle category according to the Directive
2001/116/EC of the European Communities (2002), the fuel used, the date the vehicle first entered into
service, curb weight, engine displacement and the maximum net power, where
the maximum net power is defined as the maximum value of the net power
measured at full engine load (UNECE Regulation No. 85, United Nations, 2013) and the curb
weight is the weight of the vehicle without the driver or any other
additional load (Regulation No. 540/2014 of the European Parliament, 2014).
For 2011 (the year our measurement study was conducted) we used the Eurostat
vehicle fleet statistics (for Europe and Slovenia) and the Slovenian National
Interoperability (NIO) portal (http://nio.gov.si/), where we gained detailed
information on Slovenian car fleet; and compared them to our measured fleet.
The Eurostat statistics for cars in Europe include countries that reported
not only the total number of cars but also the information on which fuel
they used and their respective engine displacements (the countries included
are listed in the Supplement Sect. S2). Of the 207 185 950 passenger
cars in-use, 61 % used gasoline fuels and 34 % used diesel.
Our vehicle classification to categories was based on that of vehicle
registration information, according to the Commission Directive 2001/116/EC
(European Communities, 2002). In Europe vehicles with more than four wheels
are organized according to their purpose to categories M, N and O, on the first
level. Category M includes vehicles for the transport of passengers,
category N comprises commercial vehicles for the transport of goods, and
category O includes trailers (and semi-trailers). Further categorization of
category M pertains to the number of passenger seats and the vehicle's
maximum allowed weight, whereas the N and O categories are further segmented
regarding the vehicle's maximum allowed weight. This classification, with
further sub categories, is then, among other things, also used for
prescribing emission standards to new vehicles. Passenger cars (category M1)
and light commercial vehicles weighing less than 1305 kg (category N1-I)
have the same emission standards, even though the corresponding Euro 1 and
Euro 2 standards came into force in different years. Light commercial
vehicles have two more categories of Emission standards: N1-II (1305–1760 kg);
and N1-III (>1760) together with N2 (light commercial
vehicles with a maximum mass exceeding 3500 kg but below 12 000 kg).
Depending on the vehicle's use, the same vehicle can be registered as an M1
or N1. Similar categorization is used in the Eurostat data. There are also
many other classifications of vehicles, that depend mostly on the purpose of
their use.
Number of vehicle types in the sampled fleet, according their assigned
categories.
Category
Vehicle type
2001/116/EC
# in our fleet sample
# missing registry information
Gasoline cars
Gasoline cars
M1
24
Diesel cars
Diesel cars
M1
51
Light goods vehicles 1
N1
17
2
Goods vehicles
Light goods vehicles 2
N2
8
2
Mini bus
M2
1
Buses
M3
6
2
Heavy goods vehicles
N3
32
15
We set up three main categories: diesel cars, gasoline cars and goods
vehicles. In the gasoline cars category we included only M1 vehicles with
spark ignition engines; in diesel cars category we included M1 cars with
compression ignition engine and light goods vehicles categorized as N1;
other vehicles categorized as N2, N3, M2 or M3 were all in the goods vehicle
category. The categorization is summarized in Table 2, where it is also indicated how it overlaps with the classification in
Directive 2001/116/EC.
For some heavy goods vehicles, buses and light goods vehicles, we were
unable to obtain the vehicle verification data (foreign vehicles and
vehicles for which we were unable to note their license plates). These
vehicles were only included in the results when more detailed information
(age, engine displacement or power) about the vehicle was not needed and the
vehicle's category could be determined solely from their visual appearance.
Thus, we kept the heavy goods vehicles and vans for which we did not have
registration information but could categorize them as N1, N2 or N3, based on
their appearance.
Emission factor measurement results
Our total vehicle fleet sample was 139 vehicles; it consisted of 75
passenger cars (M1) of which 51 were diesel and 24 gasoline cars; 6 buses
(M3); 1 mini bus (M2); 26 light goods vehicles, of which 17 were category N1
and 8 were category N2; and 32 heavy goods vehicles (N3). We were unable to
obtain the registry data for 2 buses, 4 of the light goods vehicles (2
categorized as N1 and 2 as N2), and 15 of the heavy goods vehicles (N3). The
fleet sample is summarized in Table 2.
We compared our measured fleet composition on the vehicles' age and size
with the information on the Slovenian and European vehicle fleet statistics
(Sect. 3.1). We present our results as BC, PN and NOx EF
distributions for the vehicle categories and compare them to results of
other similar studies in Sect. 3.2. We further demonstrate how the EFs of
each group depend on their age, by grouping them according to years when
Euro 3 and Euro 4 standards became effective. Even though the purpose of use
is indeed important when classifying vehicles; but with such categorization
the mechanical features may be overlooked. A single car (for example Renault
Kangoo or similar) can be classified as a personal vehicle or a light goods
vehicle. To see how mechanical and physical features of the vehicles affect
the emissions, we disregarded the purpose-based categorizations and observed
the effect of engine maximum net power, and the ratio between engine maximum
net power and vehicle mass in Sect. 3.4. In Sect. 3.5 we present the
contribution of high emitters to the sampled fleet cumulative emissions.
Comparison of sampled vehicle fleet and Eurostat data
The fleet sample size determines the representativeness of the measured
fleet. According to Ban-Weis et al. (2009), about ≥30 trucks should be
a large enough sample (presuming that the sampling was indeed random) for
the sample mean to equally likely to fall below or above the sample mean.
Our category samples were larger than the above threshold for diesel cars
and goods vehicles, and very close to the threshold for gasoline cars. This
makes us confident that the sample is large enough to be representative of
the on-road fleet during the approximate period of the campaign on East–West
and North–South trans-European corridors V and X. In order to establish the
relationship of our data as representative of the Slovenian and the average
European fleet, we used Eurostat data to compare the size and age
composition of the three investigated vehicle fleets.
In Sect. 3.1.1 we show a comparison between the European, Slovenian and
the campaign passenger car fleets (only M1 vehicles) according to the fuel
used, engine displacement and age, and in Sect. 3.1.2 the composition of
goods vehicle fleets (N1, N2 and N3) according to their size and age.
Passenger car fleets according to the fuel used and engine displacement at
the end of the year 2011.
Fleet
Total
Gasoline
Diesel
Of total
Less than
From 1400 to
2000 cm3
Of total
Less than
From 1400 to
2000 cm3
1400 cm3
1999 cm3
or over
1400 cm3
1999 cm3
or over
Europe
207 185 950
61 %
49 %
44 %
7 %
34 %
5 %
76 %
19 %
Slovenia
1 089 335*
63 %
61 %
37 %
3 %
36 %
4 %
79 %
17 %
Our fleet
75
32 %
50 %
42 %
8 %
68 %
0 %
75 %
25 %
* The Slovenian fleet in Eurostat (total vehicles 1 066 490) slightly differs
from the NIO database, which is reported in this table, but overall reports almost the same percentages of the vehicle composition.
Passenger car fleets according to their age, at the end of the year 2011.
10 years or over
From 5 to 10 years
From 2 to 5 years
Less than 2 years
Europe
Total
42 %
28 %
19 %
11 %
Slovenia
Total
39 %
34 %
18 %
9 %
Gasoline
50 %
25 %
15 %
9 %
Diesel
18 %
48 %
23 %
11 %
This study
Total
27 %
47 %
29 %
7 %
Gasoline
50 %
25 %
17 %
8 %
Diesel
16 %
49 %
29 %
6 %
Passenger cars
From Table 3 we can see that the combination of
cars in the European and Slovenian fleets are very similar. The percentage
of diesel and gasoline cars in the European fleet is 34 and 61 %,
while the Slovenian fleet has 36 and 63 % of diesel and gasoline cars,
respectively. The engine displacements of diesel or gasoline engines are
similar. Both fleets show that most gasoline cars have engine displacements
smaller than 1400 cm3 (49 and 61 % for European and Slovenian
fleet respectively) and that only a small portion of gasoline cars have an
engine displacement larger than 2000 cm3 (7 and 3 % for the
European and Slovenian fleets, respectively). Most diesel-powered cars have
an engine displacement in the size range of 1400 to 2000 cm3 (76
and 79 % respectively); the fewest have an engine displacement smaller
than 1400 cm3 (5 and 4 % respectively).
The gasoline and diesel car engine displacement segregation of the campaign
fleet is representative of the European and Slovenian fleets, where, again
most gasoline cars (50 %) had engine displacements smaller than 1400 cm3,
followed by 42 % of cars with engine displacements in the range
of 1400 to 1999 cm3 and the fewest gasoline cars had engine
displacements larger than 2000 cm3 (8 %). For diesel cars, the share
was – as in the European and Slovenian fleets – largest for 1400 to 1999 cm3
sized engines (75 %), followed by 25 % of diesel cars with
engine displacements larger than 2000 cm3. We did not encounter any
diesel cars with engine displacements smaller than 1400 cm3.
European and Slovenian car fleet statistics also compare well if segregated
by the age of the passenger cars. From Table 4 we
can see that the two have almost the same percentage in all four age groups
set by Eurostat; the largest difference between them is only 6 %. Most
passenger cars in both fleets were in use for 10 years or more (42 and
39 % for the European and Slovenian fleets respectively), followed by the
group of cars that was in use for between 5 and 10 years (28 and 34 %
respectively), almost 20 % were in use for between 2 and 5 years and about
10 % were in use for less than 2 years.
Our total measured passenger car fleet consisted of somewhat more cars in
the ages of 2 to 10 years, and fewer vehicles that were over 10 years than
were in the Slovenian and European fleets. Using the NIO database we
separated Slovenian diesel and gasoline car fleet using 10, 5 and 2 years in
use as delimiters. In Table 4 we can see that we
get almost the same percentages in all bins for both diesel and gasoline
cars in our measured fleet and the Slovenian car fleet. But because, unlike
in European or Slovenian fleet, there were more diesel than gasoline cars in
our fleet, and because half of gasoline cars were older than 10 years and
only 18 % of diesel cars were in that age group, the age of our total
fleet does not match the Slovenian or European total car fleet age
distribution.
During our measurements, our prime focus was diesel cars, because they are
commonly found in Slovenia and Europe and are the most problematic with
regard to emissions of BC and NOx. We used gasoline cars for the
control and heavy goods vehicles for comparison with previous studies that
used similar techniques. There are, therefore, a greater percentage of cars
powered by diesel (68 %) than gasoline (32 %) in the fleet of this study
than there are in Europe or Slovenia in general and therefore the age of our
total passenger car fleet does not match the total Slovenian nor European
passenger car age groups. By analyzing the age distribution within diesel
and gasoline cars separately we have shown that our two subcategories do
indeed match the Slovenian fleet from which we sampled from and are thus
representative of the Slovenian vehicle fleet, and most likely also for the
European car fleet, as the two are very similar.
Statistics on lorries weight in 2011 for Europe and Slovenia.
Total
Less than
From 1500 to
From 5000 to
10 000 kg
1500 kg
4999 kg
9999 kg
or over
Europe
17 994 007
79 %
14 %
3 %
4 %
Slovenia
75 508
71 %
14 %
7 %
8 %
Statistics on lorries age in year 2011 for Europe and Slovenia.
Total
Less than
From 2 to
From 5 to
10 years
2 years
5 years
10 years
or over
Europe
17 995 713
10 %
20 %
26 %
43 %
Slovenia
75 508
11 %
25 %
32 %
32 %
Comparison of EF with other similar on-road studies.
Study
Study type
Vehicle type
EF BC (gkg-1)
EF PN (1015 kg-1)
EF NOx (gkg-1)
Shorter et al. (2005)
Chasinga
Diesel buses
34.5 (8.1–117.1)
CRT
27.8 (±6.3)
Schneider et al. (2008)
Chasingb
HGV
0.22±0.14
8.3±5.8
18±14
Ban-Weiss et al. (2009)
Remote s.a
HGV
1.7 (0.1–20)
4.7 (0.2–40)
Dallmann et al. (2011)
Remote s.d
HGV (2009)
1.07±0.18
25.9±1.8
HGV (2010)
0.49±0.08
15.4±0.9
Dallmann (2014)
Remote s.d
HGV
0.62±0.17
Hudda et al. (2013)
Mobile
LDG
0.07±0.05
0.43±0.26
3.8±1.4
HDD I-710
0.41±0.21
4.2±3.4
15±9.2
HDD freeways
1.33±0.33
5.2±3.1
16±10
Wang et al. (2012)
Chasingc
HGV Beijing
0.4 (0.2–0.8)
47.3 (38.1–62.5)
HGV Chongqing
1.1 (0.7–1.6)
40.0 (31.7–48.1)
Carslaw and Rhys-Tyler
Remote s.e
Gasoline cars
5.6 (1.6–28.1)
(2013)
Diesel cars
16.37 (15.7–21.6)
Van (N1)
18.9 (17.6–24.7)
HGV (all)
39.8 (36.7–50.6)
EEA (2013c)
Emission inventory
Gasoline cars
8.73 (4.48–29.89)
Diesel cars
12.96 (11.2–13.88)
LGV
14.91 (13.36–18.43)
HGV
33.37 (28.34–28.29)
This study
Chasingc
Gasoline cars
0.28 (0.15–0.46)
1.95 (1.08–4.88)
6.34 (3.77–10.6)
Diesel cars
0.79 (0.36–1.36)
4.4 (2.62–9.03)
15.43 (8.82–22.63)
Goods vehicles
0.47 (0.24–0.72)
11.49 (2.55–19.76)
27.71 (17.89–38.24)
LGV (N2)
0.64 (0.37–0.96)
16.8 (8.22–19.01)
23.16 (17.89–27.46)
Buses
0.4 (0.24–0.65)
9.99 (1.91–19.23)
55.88 (39.09–55.9)
a mean (range); b mean ± standard deviation;
c median (1st and 3rd quartile); d mean ± 95 %
confidence interval; e emission ratios from Carslaw and Rhys-Tyler (2013) paper were converted
to EFs using the same molecular weights and carbon fraction as in formula 1, for
HGV we take the average of both HGV groups they report HGV(3.5–12 t) and HGV(>12 t);
presented are average values for all Euro standards in a group, in parenthesis are the smallest and largest mean value of emission standards.
Goods vehicles
The goods vehicles are much more versatile in their purpose and hence the
mass they have to carry and power they have to produce. We were able to get
the registration information for many of the sampled vehicles (28 out of 47)
to identify the technical differences between the vehicles. Below, we show
the representativeness of the Slovenian fleet for Europe. Our sample seems
big enough to be representative, given the previously published criteria
(Ban-Weis et al., 2009).
Black carbon (BC), particle number concentration (PN) and NOx
emission factor (EF) distributions for gasoline and diesel cars, light and
heavy goods vehicles. Note the EF logarithmic scale.
Eurostat does not report the number of heavy goods vehicles as N1, N2 and
N3, rather it reports the number of lorries (defined as: rigid road motor
vehicle designed, exclusively or primarily, to carry goods) by their load
capacity (defined as maximum weight of goods declared permissible by the
competent authority of the country of registration of the vehicle). The data
thus include vehicles with a gross weight of not more than 3500 kg but
excludes tow trucks. From Table 5 we can see that
lorries with load capacity less than 1500 kg are most numerous in both
Slovenian and European fleet and that the vehicles with load capacity over
10 000 kg are fewest. With Tables 5 and
6 (where we report Eurostat data for the European and Slovenian
fleet), we demonstrate that the Slovenian vehicle fleet from which we
sampled the most vehicles from is representative of European average both
regarding the size segregation and vehicle age. We could not make an
indirect comparison of Eurostat data to our sample fleet because we did not
get the load capacity reported for most of our measured vehicles, and
because the number of license plates we could collect was low. Nonetheless,
we used the NIO database and found that in the Slovenian fleet there were
72 % of N3 goods vehicles weighing less than 12 000 kg that were not road
tractors or special purpose vehicles, while in our fleet there were 57 %
of such vehicles. We binned the vehicles according to their age: those that
were in use for less than 10 years, 5 to 10, and less than 5 years. The
Slovenian fleet consisted of 38, 38 and 24 % vehicles in each
categories, respectively, while the measured vehicles consisted of 21,
50 and 29 % respectively. Here the size of the sample was only 14
vehicles for which we had registry information. The discrepancy is larger
because of the larger diversity in vehicle size among the goods vehicles
than for personal cars, and because our sample size is small.
Emission factors distributions and comparison to other studies
We determined EFs of different type vehicles, grouped them into three
categories: gasoline cars, diesel cars and goods vehicles (as described in
Sect. 2.2), and present their BC, NOx and
PN EF distributions in Fig. 2. Because the
formation paths for the three pollutants differ (see Supplement Sect. S1) and technological solutions for the three vehicle categories
differ, their EF distributions also show different tendencies. The median BC
EF for diesel cars (0.79 gkg-1) is the highest of the three vehicle
groups, followed by goods vehicles (median 0.47 gkg-1) and gasoline
cars (0.28 gkg-1), where also the lowest BC EFs are to be found. The
median of NOx EF distribution is highest for goods vehicles (27.71 gkg-1),
followed by diesel cars (15.43 gkg-1), and again lowest
for gasoline cars (6.34 gkg-1). We can observe similar trends with PN
EF distribution – highest median value for goods vehicles (11.49×1015 kg-1),
followed by diesel cars (4.4×1015 kg-1) and gasoline
cars (1.95×1015 kg-1). The shapes of the PN distributions are
different from the shapes of the NOx EF distributions. NOx EF
distributions have the narrowest range of the three investigated pollutants
for all three vehicle groups, while PN EF distributions are broad and in the
case of goods vehicles even bimodal. They would remain bimodal even if buses
and light goods vehicles (N2) would be excluded from the analysis.
In Table 7 we compare the results of our study to
other chasing and remote-sensing studies that measured the same species
(Ban-Weiss et al., 2009; Carslaw and Rhys-Tyler, 2013; Dallmann et al.,
2011; Hudda et al., 2013; Schneider et al., 2008; Shorter et al., 2005; Wang
et al., 2011, 2012). Remote-sensing studies were included because good
agreement between the results of the remote-sensing and chasing techniques
was found by Ježek et al. (2015), where it has been shown that with
multiple measurements of the same vehicle with the stationary method, we can
obtain a similar distribution as when measuring the same vehicle with the
chasing method, and that the median value of both techniques, is similar.
We did not compare our results to other study types such as tunnel
measurements, chassis dynamometer tests or measurements with portable
emission measurement systems, as they have already been discussed in other
studies (e.g. Shorter et al., 2005; Wang et al., 2012).
The BC EF median of goods vehicles we measured (0.47 gkg-1) is similar
to the mean value of HGV fleet reported by Dallmann et al. in their 2011
study after additional emission control was implemented
(0.49 gkg-1); it
compares well to the results of Wang et al. (2012), for HGVs from the Beijing
area (0.40 gkg-1), where there are also more strict emission control
standards implemented when compared to surrounding provinces and to the
results of Hudda et al. (2013), who report 0.41 gkg-1 BC EF for high
cargo route in California (I-710). While BC EFs of these studies (including
ours) agree, NOx EFs do not. While NOx EFs were high in the
Chinese study (47.3 and 40 gkg-1 for Beijing and Chongqing
respectively), they were much lower in the two US studies (∼15 gkg-1). The lower EF for the US studies may be due to a different
mix of vehicles due to promotion of the purchase of newer vehicles. The
median value of the NOx EF distribution (27.7 gkg-1) observed for
goods vehicles lies closer to the average HDV fleet value reported by
Dallmann et al. (2011) before the active replacement rule was implemented
(25.9 gkg-1), and to the results of another US study (Shorter et al.,
2005) where they report NOx EF for buses equipped with CRT (27.8 gkg-1).
The two European studies (Carslaw and Rhys-Tyler, 2013;
Schneider et al., 2008) report similar NOx EF for different vehicle
types – while Schneider et al. (2008) measured 18 trucks in Germany by
chasing them on the road, and report NOx EF of their measured fleet to
be 18 gkg-1. Carslaw and Rhys-Tyler (2013) report similar values 18.9 gkg-1
for vans (N1), but much higher for goods vehicles (average of
HGV: 37.88 gkg-1). The reason only BC or NOx EF between our
measured fleet and other studies match may be related to the different ages
of the investigated vehicle fleets. We will address this again in Sect. 3.3, where we investigate the dependency of the determined EFs to vehicle
age in their respective category.
The NOx EF values of the gasoline and diesel cars in this campaign (6.3 and 15.4 gkg-1 respectively) coincide with those reported by Carslaw
and Rhys-Tyler (2013) (5.6 and 17.1 gkg-1 respectively). The median
NOx EF of gasoline cars in this campaign is slightly lower than that
reported by the EEA (8.7 gkg-1) in Tier I approach of their guide book
(EEA, 2013c); while those of diesel cars and LDV in this campaign are
slightly higher than the NOx EFs in the aforementioned guide book (13.0 gkg-1).
The goods vehicles' NOx EFs (27.7 gkg-1) from this
campaign agree with those reported by Shorter et al. (2005) for CRT (CRT
stands for continuous regenerating technology) equipped buses (27.8 gkg-1);
and to HGV NOx EFs (25.9 gkg-1) reported by Dallmann
et al. (2011) for HGV emissions before vehicles had to be retrofitted with
additional exhaust after-treatment devices. The NOx EFs of goods
vehicles measured in the present campaign are lower than HGV NOx EFs
reported by Wang et al. (2012) (40.0 and 47.3 gkg-1), who
used the same measurement method; lower than Carslaw and Rhys-Tyler (2013)
(39.8 gkg-1) who used a stationary remote-sensing method; and lower
than HGV EF reported by EEA (33.4 gkg-1). This may indicate either
that our goods vehicles sample emitted less per unit of fuel; or that the
measurement techniques used produce different results. We have shown in
Fig. 1 how using two different integration
approaches yields in up to 16 % different results. Some differences
between the studies may arise from using the average value for
representation of the vehicle categories EF instead of the median, which is
not as strongly influenced by super emitters as the average.
The weight a truck engine has to pull can change drastically from an
unloaded truck to twice or three times its unloaded mass, therefore we would
expect its emissions would also change a lot more than we would expect them
to change with a passenger car. This is one more variable that would be
difficult to monitor under controlled condition protocols.
HGV PN EF from Ban-Weiss et al. (2009) (4.7×1015 kg-1) and from
the study of Hudda et al. (2013) (4.2 and 5.2×1015 kg-1)
coincide with those of here presented diesel cars PN EFs
(4.4×1015 kg-1); and Schneider et al. (2008), PN EF
(8.3×1015 kg-1) lie closer to our goods vehicle PN EF
median (11.49×1015 kg-1). The PN EF is most difficult
to determine and compare because it depends on the measurement instrument
and sampling conditions. Our measurements were conducted while chasing
vehicles on highways and regional roads in winter, while others measured EF
with a remote-sensing method at the end of a tunnel in summer. Each study used
different measurement instruments with a different particle size measurement
range.
BC, NOx and PN EFs according to different vehicle categories and age group
subcategories: gasoline passenger cars (GC, blue), diesel passenger cars
(DC, black), and goods vehicles (GV, red). Note the EF logarithmic scale;
same figure in linear scale can be found in Supplement Fig. S3.
Emission factors and vehicle age
In this section we have further broken down each of our four vehicle groups
to three age subgroups: less than 5 years; 5 to 10 years; and 10 or more
years in use. We wanted to observe if newer vehicles showed an improvement
in their emissions per unit of fuel burned. The 5- and 10-year limits should
roughly separate vehicles in three groups that comply with either the entry
of Euro standards 4 or 5 (less than 5-year old vehicles), Euro standard 3
(5–10-year old vehicles), and Euro 2, 1 or pre-Euro vehicles (over 10-year
old group). A clear separation between vehicles compliant Euro standards
cannot be made based solely on the date the vehicle was put in use, because
an improved vehicle may be put on the market before the date when the new
standard is enforced, or a vehicle that is compliant to the old standard may
still be put to use 1 year after the new standard enforcement date
(2001/116/EC European Communities, 2002). The vehicle age should reflect not
only the deterioration of the engine and exhaust system, but also the
technological advances made in engines and exhaust systems over the years
due to stricter emission standards.
The results show some improvement for the three investigated pollutants
(Figs. 3 and S3). For BC EFs the improvement is
most evident for less than 5-year old diesel cars, where we can see a 60 %
drop in median values from 5–10-year old diesel cars to those with age
less than 5 years. This reduction most probably reflects the impacts of
regulations to reduce the PM vehicle emissions from Euro 3 to Euro 4 by
50 %. The reduction was probably achieved with the increased use of diesel
particle filters (DPF), which are commonly used in the post Euro 5 cars. We
can also observe a 55 % decrease in median BC EF of gasoline cars from the
oldest (10 or older) to the newest group (5 years or less). These vehicles
are less critical regarding PM emissions than diesel cars. However, due to
increased PM emissions, and especially PN emissions, of direct injection gasoline
cars, both of these parameters are limited in recent Euro emission standards. Our results show that
compared to the newest diesel car category the gasoline cars have lower BC EF
medians in all three age groups. We can observe a 41 % decrease in BC EF
median from goods vehicles older than 10 years to the 5–10-year category.
Worryingly, the newest goods vehicles median BC EF increased by 34 % in
comparison to the 5–10-year old group. Emission standards from Euro III to
Euro IV for goods vehicles demanded PM emissions (in gkW-1h-1) to
reduce 5 fold. Unlike passenger cars, the emission reduction of goods
vehicles was achieved with SCR and not with DPF, and thus the soot emissions
were not limited as efficiently.
In Fig. 3 (and Fig. S3) we observed a 67 % decrease
in goods vehicles PN EFs (in 1015 kg-1) from 5–10-year old
vehicles to those that were in use for less than 5 years. This may indicate
that either more agglomerated soot particles were being emitted or emissions
of some of the particulate precursors had been reduced. Median PN EFs
reduced by 67 % from the oldest to the newest diesel car group. For
gasoline cars the PN EFs varied the most within individual age groups, where
individual vehicles with high emissions skewed the distribution.
In Fig. 3 (and Fig. S3) we can observe the gradual
decrease of NOx EFs from gasoline cars to diesel cars to goods
vehicles, as it is also shown in Fig. 2, where
also vehicles for which we did not get more detailed information were
included. Goods vehicle NOx EFs are showing an appreciable decrease in
average and median values from the oldest to newest age group (50 and 70 %
respectively), which we postulate is due to increased use of SCR in newer
post Euro V vehicles, which can effectively reduce NOx emissions. When
separated by age, we can see that now both NOx and BC EF correlate
better to some of the previously published studies
(Table 7). The 10 year or older goods vehicles (BC
and NOx EF respectively 0.7, 43.95 gkg-1, please see
Figs. 3 and S3) relate better with Wang et al. (2012), Chongqing EFs; and
our less than 5-year old goods vehicles (median BC and NOx EF,
respectively 0.55, 13.37 gkg-1) relate better with the
most recent situation reported in the US for high cargo routes in California
(I-710) by Hudda et al. (2013).
Diesel cars' maximum NOx EFs increased in the newest group but the
median of the group decreased by 24 % in comparison to 5–10-year old
diesel cars. NOx emission standards for diesel and gasoline cars were
introduced with the Euro 3 standard. We could observe a reduction of
gasoline car median NOx EF from those in use for over 10 years to those
in use for 5–10 years. At this time the use of the three-way catalysts was
common in the market and according to our results efficient in reducing
NOx emissions. The median did not reduce further for diesel cars that
were in use for less than 5 years but the average value did. The decrease of
emissions is smaller than we would expect it to be according to the newer
European emission standards. We postulate that this is because the emissions
of Euro 5 diesel cars were achieved with DPF, not including de-NOx
devices, in such instances driving that would be more aggressive than NEDC
would not reflect more stringent NOx Euro emission standards in
real-world driving. In the study of Carslaw and Rhys-Tyler (2013), they found
a satisfactory reduction of average NOx EF only for gasoline cars but
not for diesel cars.
Carslaw and Rhys-Tyler found an influence of vehicle manufacturer on
NOx EFs for Euro 4/5; this could potentially explain the skewed
NOx EF distribution observed in our fleet, if some of the manufacturers
would be disproportionally represented. However, Carslaw and Rhys-Tyler did
not reveal the brands that produce lower EF values; and our sample size is
too small compared to the number of manufacturers for us to consider
debating such trends in our fleet.
The reason the EF distributions are skewed and some an order of magnitude
higher than the rest may be because, at the time of our measurements, these
cars were somehow compromised, e.g., not well maintained, or frequently
operating in transient conditions that favored high-pollutant emissions.
On-road measurements of individual in-use vehicle fleets can provide useful
information about the fleet emissions by exactly including such vehicles.
BC, NOx and PN Efs according to engine power (left) and power per mass (right); red boxes
present gasoline engines (GE) and black boxes present all diesel engines
(DE) regardless of their vehicle category. Note the EFs are on logarithmic
scale; same figure in linear scale can be found in Supplement
Fig. S4.
Cumulative distribution of all vehicle emissions. Fractions of vehicles
are distributed from highest to lowest emitting vehicles. The result shows
that 10 % of vehicles contribute about a half of total BC and NOx
emissions.
Emission factors according to maximum net engine power and maximum net
engine power to vehicle weight ratio
In addition to the information about the vehicle engine type, their category
and the date of first use, the registration database also provided
information about the engine's maximum net power and vehicle curb weight. We
present in this section the EFs sorted according to the engine maximum net
power and the ratio of engine's maximum net power to vehicle's curb weight.
Here, we do not use the same vehicle groups as in the previous subchapter.
Rather we separated the vehicles into gasoline and diesel engines and then
further according to different size bins for both engine maximum net power
and maximum net power to weight ratio. The sizes of the bins were determined
in a way that a single bin size would not include a disproportionally large
number of vehicles and that each bin would have enough vehicles for a
statistical presentation. There are also some gaps between the adjacent
bins; this is because there were no vehicles in that range. The results are
shown in Figs. 4 and S4.
When EF are sorted by a vehicle's engine maximum net power, we can see that
diesel engines in the lowest maximum net power bin (less than 70 kW)
feature highest median BC EFs and that the more powerful diesel engines feature
lower BC EFs. The trend is reversed for NOx EF, where more powerful
larger vehicles feature higher NOx EF. There is an exception for
NOx EF in the least powerful diesel group, which feature relatively
high NOx EF compared to the adjacent engine power bins.
The ratio of maximum engine power to vehicle curb weight can give a rough
estimate of the engine load under which the vehicle has to operate in normal
driving conditions. Large trucks have high vehicle mass but low maximum net
power to vehicle mass ratio. Smaller vehicles have smaller mass but higher
maximum net power to vehicle mass ratios, and for the smallest vehicles the
ratio again decreases. A vehicle with lower maximum net power to mass ratio
driven in similar driving conditions and with a similar driver behavior
would have its engine operating at higher loads leading to higher
in-cylinder temperatures. Operation at higher in-cylinder temperatures would
result in more thermic NOx. This trend in NOx can be observed in
Figs. 4 and S4 for both diesel and gasoline
engines, where we can see that vehicles with low power to mass ratio produce
higher NOx EF and vehicles with high power to mass ratio produce lower
NOx EF. For BC and PN EF the trend is not as clear as it is for
NOx, it could be described as a gradual increase of EF from low to high
power to mass ratios but in the highest power to mass ratio bin the median
BC and PN EF drop.
We separated the gasoline vehicles into two groups for each observed
parameter. The differences between gasoline vehicle categories are difficult
to observe. We postulate this is because we were only operating with cars
and the change in the vehicle mass and mass to power ratio was smaller than
it was for the vehicles with diesel engines which included trucks.
Contribution of high emitters to the measured fleet
The contribution of high emitters to the measured vehicle fleet was
calculated as cumulative emissions. To exclude large differences in fuel
consumption between trucks and cars, we calculated high emitter contribution
separately for goods vehicles, gasoline cars and diesel cars. The cumulative
emission distribution of our vehicle fleet was calculated for vehicles from
highest to lowest emitters as it was previously done in similar studies
(Ban-Weiss et al., 2009; Dallmann et al., 2012; Wang et al., 2011, 2012).
The results in Fig. 5 show that 25 % of the highest-emitting vehicles in each vehicle category produce 50 to 65 % of BC
emissions, 47 to 55 % of NOx emissions and 61–87 % of PN
emissions. The high contributions of super emitters are the statistical
cause of the non-symmetrical distributions and are responsible for the
mismatch between the median and the average EF values. Excluding high
emitting vehicles or improving their emission rates by retrofitting them
with additional after treatment devices, such as the case in the Port of
Oakland, US (Dallmann et al., 2011) can decrease traffic emissions.
Conclusions
During the measurement campaign the BC, PN and NOx EFs for 139
different vehicles were successfully determined. The sample fleet statistics
were compared to Eurostat data for Slovenia and Europe. An excellent
agreement between the composition of the average European and Slovenian car
fleet, and the car sample fleet sampled in this campaign was found. The main
results of this research are the first reported on-road BC EF for diesel
cars, and first BC, PN and NOx EF for passenger cars measured with the
on-road chasing technique. In order to compare the results of this study to
previous ones, EFs of goods vehicles were also determined. EF distributions
for BC, PN and NOx were presented for three vehicle groups: diesel
cars, gasoline cars and goods vehicles. Differences between the EF frequency
distributions of the three vehicle categories for all three investigated
pollutants were observed, the most important being the highest median BC EF
value of diesel cars, and an increase in NOx EFs from the least
powerful to more powerful diesel vehicles.
The results of this study were compared to the results of previous studies
that used similar methods. The median BC EFs for the diesel cars (0.79 gkg-1)
determined in this study is similar to the HGV EFs mean reported
by Dallmann et al. (2011) and Wang et al. (2012), where vehicles were subject
to less strict emission regulations. The goods vehicles' BC EF median determined
here resembles the EFs determined for vehicles subject to stricter emission
standards. The goods vehicle median NOx EF reported in this study
resembles those determined by Dallmann et al. (2011) and Shorter et al. (2005).
The median BC EF value of newer diesel and gasoline cars (less than 5 years)
is lower than the value for the older car categories. For the goods vehicles
it lies between the medians of the two older goods vehicles groups. Contrary
to BC EF, goods vehicles showed a significant 73 % decrease in the
NOx EF median values from vehicles that were in use for over 10 to
those in use for less than 5 years. We postulate this is because different
after-treatment approaches were used for passenger cars and goods vehicles.
PN EF median values decreased for vehicles in use for less than 5 years in
all three vehicle groups compared to older ones, but unfortunately the span
of PN EFs of goods vehicles and gasoline cars increased. We attribute the
decreases to advances made in engine operation and exhaust after treatment
devices.
The contribution of highly emitting vehicles was calculated and, as in
previous studies (e.g. Ban-Weiss et al., 2009; X. Wang et al., 2015; J. M. Wang
et al., 2012), a small number of vehicles (25 %) was found to
disproportionately contribute to the total fleet emissions (47 to
87 %). The exclusion of high emitters by retrofitting old vehicles with
after-treatment devices and encouraging the sale of new vehicles through the
exchange of older vehicles, has shown to be an effective measure to reduce
vehicle emission rates (Dallmann et al., 2011) locally. Unfortunately, the
older vehicles might be sold in countries beyond the reach of the EU
regulations, and would still have a negative impact on air quality and the
climate elsewhere.
The methodology used in this study is a relatively simple and efficient
approach for monitoring emissions of the in-use vehicle fleet, and
investigating the effectiveness of emission reduction measures (also shown
in Dallmann et al., 2011; Wang et al., 2011). Real-world measurements
are important because individual vehicle emissions depend not only on the
vehicle type approval at the time it is put on the market, but are also on
their maintenance and the driving conditions.