Particle and gaseous emissions from individual diesel and CNG buses 5339

Abstract. In this study size-resolved particle and gaseous emissions from 28 individual diesel-fuelled and 7 compressed natural gas (CNG)-fuelled buses, selected from an in-use bus fleet, were characterised for real-world dilution scenarios. The method used was based on using CO 2 as a tracer of exhaust gas dilution. The particles were sampled by using an extractive sampling method and analysed with high time resolution instrumentation EEPS (10 Hz) and CO 2 with a non-dispersive infrared gas analyser (LI-840, LI-COR Inc. 1 Hz). The gaseous constituents (CO, HC and NO) were measured by using a remote sensing device (AccuScan RSD 3000, Environmental System Products Inc.). Nitrogen oxides, NO x , were estimated from NO by using default NO 2 /NO x ratios from the road vehicle emission model HBEFA3.1. The buses studied were diesel-fuelled Euro III–V and CNG-fuelled Enhanced Environmentally Friendly Vehicles (EEVs) with different after-treatment, including selective catalytic reduction (SCR), exhaust gas recirculation (EGR) and with and without diesel particulate filter (DPF). The primary driving mode applied in this study was accelerating mode. However, regarding the particle emissions also a constant speed mode was analysed. The investigated CNG buses emitted on average a higher number of particles but less mass compared to the diesel-fuelled buses. Emission factors for number of particles (EF PN ) were EF PN, DPF = 4.4 ± 3.5 × 10 14 , EF PN, no DPF = 2.1 ± 1.0 × 10 15 and EF PN, CNG = 7.8 ± 5.7 ×10 15 kg fuel −1 . In the accelerating mode, size-resolved emission factors (EFs) showed unimodal number size distributions with peak diameters of 70–90 nm and 10 nm for diesel and CNG buses, respectively. For the constant speed mode, bimodal average number size distributions were obtained for the diesel buses with peak modes of ~10 nm and ~60 nm. Emission factors for NO x expressed as NO 2 equivalents for the diesel buses were on average 27 ± 7 g (kg fuel) −1 and for the CNG buses 41 ± 26 g (kg fuel) −1 . An anti-relationship between EF NO x and EF PM was observed especially for buses with no DPF, and there was a positive relationship between EF PM and EF CO .


Introduction
It is acknowledged that combustion processes, especially traffic-related emissions, contribute significantly to total particulate air and gaseous pollutants in urban environments.Many epidemiological studies have shown that particles have adverse health effects (Pope and Dockery, 2006).Particles also have an effect on climate either directly via scattering and absorption of radiation or indirectly via its influence on the formation of clouds.
When measuring particle emissions, mass basis is often used.This implies that such data are dominated by large particles.Numerically vehicle exhaust is dominated by ultrafine particles (UFPs), i.e. particles with a diameter < 100 nm (Janhall et al., 2004;Harrison et al., 1999;Kumar et al., 2010).Therefore an alternative way of presenting particle emissions is needed -i.e.looking at the number of particles emitted -to enable accounting for the small particles that on a mass basis are negligible.Further, health risks are probably dominated by the UFPs (Donaldson et al., 1998;Delfino et al., 2005;Valavanidis et al., 2008).Thus, there is an obvious need to ascertain the emission of particles from traffic regarding number and size in order to establish effective air quality management strategies.Particles measured in close vicinity of the emission source are primary, i.e. emitted as particles from the tailpipe, or secondary, i.e. formed during the expansion and cooling of the hot exhaust gases.The former are often in the form of agglomerates of solid phase material, whereas the latter are more volatile (Morawska et al., 2008).Additionally, traffic contributes to the formation of secondary organic aerosols (SOAs); however, the magnitude of this contribution is very uncertain (Robinson et al., 2007).This is a chemically induced particle formation (time scales of hours to days) which is very important on a regional and global scale (Hallquist et al., 2009).

Published by Copernicus
The particle emissions from any combustion source can be derived from the emission ratio of the particle concentration to a co-emitted trace gas, such as CO 2 or NO x (Janhall and Hallquist, 2005).Knowing the emission factor for the chosen trace gas (EF gas ), an emission factor for particle number (EF PN ) or mass (EF PM ) can be estimated (Hak et al., 2009).
where part and gas are measured changes in the concentration of particle number/mass and trace gas, respectively.Alternative ways of measuring particle emissions from vehicles are at the kerbside, often giving values for the average fleet, or by chassis dynamometer, measuring vehicles individually (e.g.Janhall et al., 2004;Wang et al., 1997;Ban-Weiss et al., 2010).However, in the latter case it is difficult, if not impossible, to accurately mimic the real-world dilution.
Additionally, there are chase-car experiments where the test vehicle is followed by an instrumented vehicle (e.g.Pirjola et al., 2004;Vogt et al., 2003).A challenge with this method is to avoid being influenced by other vehicles as well as keeping the distance between the target vehicle and the chasing vehicle constant.Knowledge about emissions from the on-road fleet under real-world conditions is crucial.In a recent study, EF PN was measured at the kerbside for individual vehicles for real-world dilution (Hak et al., 2009).
Along with particles, nitrogen oxides, NO x , are depicted as being the most problematic pollutant from internal combustion engines (Lopez et al., 2009).In order to meet the lower NO x and particle emission levels introduced for heavy duty vehicles (HDVs), exhaust gas after-treatment has become necessary.To reduce particle emissions from HDVs, diesel particulate filters (DPFs) are widely used.An example of after-treatment technology to reduce NO x is selective catalytic reduction (SCR), which can be found in power plants, ships and lately also in HDVs.The most common method is SCR with urea injection due to urea's low toxicity and ease in handling, but direct injection of NH 3 can also be used.In the SCR system the urea/water mixture (e.g.AdBlue®) is first added to the exhaust gas which becomes hydrolysed to NH 3 and CO 2 .In the SCR catalyst section NH 3 reacts with NO x to form N 2 and H 2 O. Another common approach to reduce NO x emissions is exhaust gas recirculation (EGR).By keep-ing a low combustion temperature and low oxygen content the formation of NO x is unfavourable; this can be achieved by recirculating a small fraction of the exhaust gas back to the cylinders.
Emissions from new HDVs in Europe are regulated by Euro standards.Currently in force since 2008 is the Euro V standard, and the Euro VI standard will be implemented in 2013.Enhanced Environmentally Friendly Vehicle, EEV, is a voluntary environmental standard which requires lower emission levels than Euro V.It was introduced together with the Euro IV and Euro V emission standards as an incentive to develop vehicles with even lower emission levels than required by regulations, and is mostly applicable to CNG heavy duty vehicles.
In order to meet the challenges with increased transportation, decreased oil resources and enhanced greenhouse gas emissions, the European Union has decided on a 10 % substitution of traditional fuels in the road transport sector (petrol and conventional diesel) by alternative fuels by the year 2020.However, the emissions from vehicles using alternative fuels have to be thoroughly studied to avoid introduction of air pollutants that can have severe health/environmental effects or other so far unknown effects or, alternatively, to establish the advantages from using these fuels.
In the literature there are some studies that have compared the particle emissions from diesel-fuelled and CNG-fuelled buses (Jayaratne et al., 2008(Jayaratne et al., , 2009;;Wang et al., 1997;Ullman et al., 2003;Lanni et al., 2003;Norman et al., 2002;Clark et al., 1999).This study takes these investigations further by determining both gaseous (NO x , CO and HC) and size-resolved particle emission factors for CNG and diesel buses belonging to different Euro classes with various aftertreatment equipment, i.e.EGR and SCR, for real-world dilution scenarios.

Experimental method
In this study particle and gaseous emissions from individual vehicles were determined by measuring the concentration change in the diluted exhaust plume compared to the concentrations before the passage and relative to the change in CO 2 concentration.By this method it is not necessary to measure absolute concentrations as the relation to CO 2 is assumed to be constant during dilution (Jayaratne et al., 2005(Jayaratne et al., , 2010;;Canagaratna et al., 2004;Shi et al., 2002;Hak et al., 2009).In addition, this method enables deriving size-resolved EFs (Janhall and Hallquist, 2005).
In total 35 different buses were studied, 28 diesel buses and 7 CNG buses.A summary of their technical characteristics including fuel used, Euro class, after-treatment system, year taken into service and kilometres travelled is shown in Table 1.
The measurements were performed at five different locations in connection to the bus depots with limited influence from other traffic.Each bus passed the remote sensing and EEPS instrumentation in two driving modes: (1) acceleration from standstill to about 20 km h −1 , and (2) constant speed of about 20 km h −1 .Before the buses were measured they were driven a distance, assuring the engines to be fully warmed up.Each bus was tested at least three times, but often more repetitions were performed.

Particle sampling
The sampling of the particle emissions was conducted according to Hak et al. (2009), i.e. an extractive sampling of the passing bus plumes where the sample was continuously drawn through a cord-reinforced flexible conductive tubing.The particles were measured with an EEPS (Engine Exhaust Particle Sizer Spectrometer, Model 3090, TSI Inc.).With this instrument, particle size distributions both regarding mass and number can be obtained in the size range of 5.6-560 nm and with a time resolution of 10 Hz.When determining the mass of particles emitted, spherical particles with unit density were assumed.The CO 2 concentration was measured with a non-dispersive infrared gas analyser (LI-840, LI-COR Inc.) with a time resolution of 1 Hz (Fig. 1).
In order to prevent the influence of the ambient temperature on the measurements for the different measurement days, the extracted sample flow was heated to 298 K before the analysis using a thermodenuder (TD; Dekati).Sizedependent aerosol losses within the TD were accounted for (user manual).

Gas sampling
The gaseous constituents NO, HC and CO were measured by using a remote sensing device (AccuScan RSD 3000, Environmental System Products Inc.).This equipment was set up with a transmitter and a receiver on one side of the passing lane and a reflector on the other (Fig. 1).The principle of this instrument has been described in detail elsewhere (Burgard et al., 2006) and will only be briefly presented here.This instrumental set-up generates and monitors a co-linear beam of IR and UV light emitted and reflected.Concentrations are determined relative to the concentration of CO 2 with a time resolution of 100 Hz.For detecting CO, HC and CO 2 the absorptions in the IR region at 2150 cm −1 , 2970 cm −1 and 2350 cm −1 , respectively, are used.For NO the absorption in the UV region at 227 nm is used.The instrumental noise of the used RSD 3000 unit was estimated with the method described in Burgard et al. (2006) using a dataset from an earlier remote sensing study, comprising more than 20 000 on-road emission measurements on passenger cars.The detection limits were then estimated as three times the standard deviation of the noise and were determined to be 18 g (kg fuel) −1 , 14 g (kg fuel) −1 and 5 g (kg fuel) −1 for CO, HC and NO, respectively.
Calibrations were conducted every 1.5-2 h of measurements by using a certified gas mixture containing 1510 ppm propane, 1580 ppm NO, 1600 ppm NO x , 3.00 % CO and 12.8 % CO 2 in N 2 (AGA Gas).The gaseous data was retrieved from the RSD system as ppm or %.

Calculation of emission factors (EFs)
Particle emission factors were derived by assuming the CO 2 concentration to be directly proportional to the fuel consump-tion, hence assuming complete combustion.For the gaseous constituents also the measured HC and CO were accounted for.In the calculations a carbon fraction of 0.865 and 0.749 for diesel and CNG fuel, respectively, was used.In this study the emission factors are presented as mass or number per kg fuel used.The gaseous pollutant emission factor for each compound (CO, HC or NO) per kilogram of fuel burnt was for diesel-fuelled vehicles calculated by applying Eq. ( 2) (Burgard et al., 2006) and for CNG-fuelled vehicles by applying Eq. (3): where CF Fuel is the carbon mass fraction of the fuels, M gas and M C are the molar mass of CO, HC, NO and C, respectively, and SF is a scaling factor.The RSD unit is calibrated with propane, and the hydrocarbons in the exhaust gas from diesel vehicles are assumed to be similar to the calibration gas, hence the molar mass of propane was used as M HC in Eq. ( 2).In Eq. ( 3) the molar mass of methane was used as this is the major constituent of CNG.The scaling factor is only applicable for determining HC; for all the other gaseous compounds SF is equal to 1.An SF is used to compensate for the known difference between non-dispersive infrared (NDIR)-based measurements and flame ionization detector (FID)-based measurements, a factor of 2 for diesel-fuelled vehicles (Singer et al., 1998) and a factor of 4.3 for CNGfuelled vehicles (Stephens et al., 1996;Singer et al., 1998).
The factor of 6 in Eq. (2) arises from the carbon atoms per molecule of propane multiplied with the scaling factor of 2. Since the remote sensing device measures NO and not NO 2 , the reported NO x emission factors have been estimated from measured NO and the default NO 2 /NO x ratios from the HBEFA 3.1 road vehicle emission model (HBEFA3.1,2010); see Table 2.The NO x emission factors were calculated by using Eq.(4): where EF NO is expressed as grams of equivalent NO 2 per kg fuel.Reporting NO x emissions as equivalent NO 2 complies with HDV emission standards (Shorter et al., 2005).
In order to be able to compare with studies expressing EFs in mass/number per km, the EFs in this study were re-calculated by using the average fuel consumption reported for the tested diesel and CNG buses, 0.38 L km −1 and 0.735 Nm 3 km −1 , respectively.For the calculations a density of 0.815 kg dm −3 and 0.70 kg m −3 was assumed (Swedish Environmental Protection Agency, 2013).These EFs (in  number/mass km −1 ) will be a lower limit as the fuel consumption during acceleration is expected to be higher.

Modelling
The measured EFs (both particles and gaseous) were also compared to modelled EFs by using the HBEFA 3.1 (2010).This model provides EFs in g km −1 for six main categories of road vehicles: passenger cars, light duty vehicles, heavy goods vehicles, urban buses, coaches and motorcycles (including mopeds).These main categories are further divided into size classes, type of fuel and emission standards.For all Euro IV and Euro V HDVs the model provides EFs separately for vehicles with SCR and for vehicles with EGR.For the class urban buses EFs are also provided for vehicles both with and without DPF.Furthermore, the emission factors are given for a large number of traffic situations based on emis-sion measurements according to different sets of real-world driving cycles (HBEFA3.1,2010).
The measured EFs in this work were compared to modelled data for a standard urban bus (15-18 tons).The driving pattern was classified according to the HBEFA 3.1 traffic situation scheme as urban access road with a posted speed of 30 km h −1 and with stop-and-go traffic.The stop-and-go traffic flow is defined as a driving cycle including many accelerations from standstill which was considered to be the driving pattern that best described the driving pattern used in the present measurements for the accelerating mode.All EFs were recalculated from g km −1 to g kg −1 by using the specific fuel consumption given in HBEFA 3.1.Used emission factors, fuel consumption and NO 2 to NO x ratios are presented in Table 2.

Emission signal
An example of typical signals in number of particles and CO 2 concentration during a bus passage is shown in Fig. 2. In this figure three successive bus passages for the same vehicle are displayed for the accelerating mode.The shape of the CO 2 peak is broader than the particle peak, which is due to the use of a small volume before the CO 2 analyser, extending the time available for the instrument to process the gas sample in order to prevent concentration peaks out of the instrument's measurement range.In Table 3 the measurement results for all the tested buses are presented.Generally there is higher variation in the data for the constant speed mode tests compared to the accelerating mode tests, which is primarily due to difficulties for the drivers to keep the same www.atmos-chem-phys.net/13/5337/2013/Atmos.Chem.Phys., 13, 5337-5350, 2013 constant speed/rpm while passing the measurement equipment on repeated occasions.However, vehicles identified as high-emitters in the accelerating mode were also generally identified as high-emitters in the constant speed mode (Table 3).

EF part for different Euro classes
In Fig. 3 the derived EF PN and EF PM for each Euro class are shown for the accelerating mode.Generally, higher EFs were obtained for buses without DPFs regarding both number and mass of particles emitted.The CNG buses emitted on average a higher number of particles compared to the diesel-fuelled buses, which is in line with previous studies (Jayaratne et al., 2008(Jayaratne et al., , 2010)).When comparing the average EF PN of the investigated diesel-fuelled buses with the CNG-fuelled buses for the accelerating mode, the EF PN for CNG buses were about five times higher (1.6 ± 0.7 × 10 15 vs. 7.8 ± 5.7 × 10 15 (kg fuel)  using the same fuel C-content assumption as in this study.However, in the case of mass of particles, the emissions from the CNG buses were on average lower compared to diesel buses.
Figure 3 also shows that a diesel bus with DPF for the accelerating mode emits on average 5 times less than a diesel bus without DPF regarding number of particles and 3 times less regarding mass of particles (4.4 ± 3.5 × 10 14 vs. 2.1 ± 1.0 × 10 15 kg −1 and 206 ± 175 vs. 696 ± 398 mg kg −1 ).
Regarding number of particles, only buses without DPF were having EFs above the average EF of all tested vehicles (see Fig. 3).The largest scatter in EF PN was, however, obtained for the CNG-fuelled buses.Out of the 15 highest PN-emitting buses, there were five gas buses (in total 7 CNG buses were tested) and 13 had no DPF installed.Regard-ing mass of particles, vehicles emitting above the average EF PM of all tested buses belonged to all Euro classes, except for buses representing Euro V with EGR and the CNGfuelled buses.The 15 highest PM-emitting buses were only diesel-fuelled buses; 12 had no DPF and four of the total five tested Euro IV with EGR buses were among these vehicles.The higher masses obtained for EGR-equipped buses without DPF may be due to the decrease in oxygen content when some of the exhaust gas is re-circulated, which favours soot formation (Seinfeld and Pandis, 1998;Maricq, 2007).
For comparison, modelled values of EF PN and EF PM using the HBFA 3.1 model are shown in Table 2.The modelled values are generally significantly lower than the measured values.A possible explanation for this can be different driving modes, acceleration versus route, including start and stops but also constant speed mode.As indicated by  using the same fuel C-content assumption as in this study.However, in the case of mass of particles, the emissions from the CNG buses were on average lower compared to diesel buses.
Figure 3 also shows that a diesel bus with DPF for the accelerating mode emits on average 5 times less than a diesel bus without DPF regarding number of particles and 3 times less regarding mass of particles (4.4 ± 3.5 × 10 14 vs. 2.1 ± 1.0 × 10 15 kg −1 and 206 ± 175 vs. 696 ± 398 mg kg −1 ).
Regarding number of particles, only buses without DPF were having EFs above the average EF of all tested vehicles (see Fig. 3).The largest scatter in EF PN was, however, obtained for the CNG-fuelled buses.Out of the 15 highest PN-emitting buses, there were five gas buses (in total 7 CNG buses were tested) and 13 had no DPF installed.Regard-ing mass of particles, vehicles emitting above the average EF PM of all tested buses belonged to all Euro classes, except for buses representing Euro V with EGR and the CNGfuelled buses.The 15 highest PM-emitting buses were only diesel-fuelled buses; 12 had no DPF and four of the total five tested Euro IV with EGR buses were among these vehicles.The higher masses obtained for EGR-equipped buses without DPF may be due to the decrease in oxygen content when some of the exhaust gas is re-circulated, which favours soot formation (Seinfeld and Pandis, 1998;Maricq, 2007).
For comparison, modelled values of EF PN and EF PM using the HBFA 3.1 model are shown in Table 2.The modelled values are generally significantly lower than the measured values.A possible explanation for this can be different driving modes, acceleration versus route, including start and stops but also constant speed mode.As indicated by  using the same fuel C-content assumption as in this study.However, in the case of mass of particles, the emissions from the CNG buses were on average lower compared to diesel buses.
Figure 3 also shows that a diesel bus with DPF for the accelerating mode emits on average 5 times less than a diesel bus without DPF regarding number of particles and 3 times less regarding mass of particles (4.4 ± 3.5 × 10 14 vs. 2.1 ± 1.0 × 10 15 kg −1 and 206 ± 175 vs. 696 ± 398 mg kg −1 ).
Regarding number of particles, only buses without DPF were having EFs above the average EF of all tested vehicles (see Fig. 3).The largest scatter in EF PN was, however, obtained for the CNG-fuelled buses.Out of the 15 highest PN-emitting buses, there were five gas buses (in total 7 CNG buses were tested) and 13 had no DPF installed.Regarding mass of particles, vehicles emitting above the average EF PM of all tested buses belonged to all Euro classes, except for buses representing Euro V with EGR and the CNGfuelled buses.The 15 highest PM-emitting buses were only diesel-fuelled buses; 12 had no DPF and four of the total five tested Euro IV with EGR buses were among these vehicles.The higher masses obtained for EGR-equipped buses without DPF may be due to the decrease in oxygen content when some of the exhaust gas is re-circulated, which favours soot formation (Seinfeld and Pandis, 1998;Maricq, 2007).
For comparison, modelled values of EF PN and EF PM using the HBFA 3.1 model are shown in Table 2.The modelled values are generally significantly lower than the measured values.A possible explanation for this can be different driving modes, acceleration versus route, including start and stops but also constant speed mode.As indicated by Table 3, EF PN/PM was generally lower for constant speed mode compared to acceleration.Modelled EF PN was the lowest for www.atmos-chem-phys.net/13/1/2013/Atmos.Chem.Phys., 13, 1-14, 2013 using the same fuel C-content assumption as in this study.However, in the case of mass of particles, the emissions from the CNG buses were on average lower compared to diesel buses.Figure 3 also shows that a diesel bus with DPF for the accelerating mode emits on average 5 times less than a diesel bus without DPF regarding number of particles and 3 times less regarding mass of particles (4.4 ± 3.5 × 10 14 vs. 2.1 ± 1.0 × 10 15 kg −1 and 206 ± 175 vs. 696 ± 398 mg kg −1 ).
Regarding number of particles, only buses without DPF were having EFs above the average EF of all tested vehicles (see Fig. 3).The largest scatter in EF PN was, however, obtained for the CNG-fuelled buses.Out of the 15 highest PN-emitting buses, there were five gas buses (in total 7 CNG buses were tested) and 13 had no DPF installed.Regarding mass of particles, vehicles emitting above the average EF PM of all tested buses belonged to all Euro classes, except for buses representing Euro V with EGR and the CNGfuelled buses.The 15 highest PM-emitting buses were only diesel-fuelled buses; 12 had no DPF and four of the total five tested Euro IV with EGR buses were among these vehicles.The higher masses obtained for EGR-equipped buses without DPF may be due to the decrease in oxygen content when some of the exhaust gas is re-circulated, which favours soot formation (Seinfeld and Pandis, 1998;Maricq, 2007).
For comparison, modelled values of EF PN and EF PM using the HBFA 3.1 model are shown in Table 2.The modelled values are generally significantly lower than the measured values.A possible explanation for this can be different driving modes, acceleration versus route, including start and stops but also constant speed mode.As indicated by Table 3, EF PN/PM was generally lower for constant speed mode compared to acceleration.Modelled EF PN was the lowest for CNG buses and highest for diesel buses, whereas the opposite was found in this study.A reason for this can be that the particle number emissions that the HBEFA model is based on often follow the PMP protocol, involving heating the particle sample to 300 • C, and the CNG particles are suggested to be volatile (Jayaratne et al., 2012).
www.atmos-chem-phys.net/13/5337/2013/Atmos.Chem.Phys., 13, 5337-5350, 2013 For the constant speed mode higher EFs were also generally obtained for buses without DPF.However, too few CNG buses were analysed in this driving mode to make a comparison between EFs for CNG buses and diesel buses.
Table 4 is a summary of the average EF PN and EF PM for diesel buses with and without DPF and for CNG buses obtained in this study (recalculated to km −1 ) and a comparison to other studies.Generally, the average EFs obtained for number of particles are within the reported ranges for diesel buses but somewhat higher for the CNG-fuelled buses.The average EF PM measured for diesel buses in this study are also within the ranges reported in other studies.In Table 4 most EF PM data is for larger particle size ranges.However, as most particles related to road traffic combustion are below 560 nm, as is shown in Figs. 4 and 5, the particle size range used in this study is comparable to PM 10 and PM 2.5 .However, important to note is that road measurements of PM 10 and PM 2.5 can include non-combustion-related particle emissions, e.g.re-suspension, and can hence be higher.It is a large variation in the reported data regarding the mass emitted for CNG buses and the data reported in this study are similar to results by Jayaratne et al. (2009) and Nylund et al. (2004).
In Lopez et al. (2009) a Euro IV diesel-fuelled bus equipped with EGR and DPF and a Euro IV diesel-fuelled bus equipped with SCR were analysed for a full driving cycle for which EF PM were determined to be 49 ± 1 and 73 ± 4 mg vehicle −1 km −1 , respectively.In this study no Euro IV with SCR were studied, but Euro V were studied, and the average EF PM for these buses (when excluding one extreme) was 68 ± 11 mg vehicle −1 km −1 .Two Euro IV diesel-fuelled buses equipped with EGR and DPF were tested: one gave similar EF PM to Lopez et al. (2009), 55 mg vehicle −1 km −1 , and the other significantly higher EF PM , 201 mg vehicle −1 km −1 .
The data presented in this study (Table 3) is a reflection of the true variation in an in-use regional bus fleet, where the variation found between similar buses (e.g.regarding fuel type and after-treatment technology) within the same Euro class can be due to engine specifics, maintenance and malfunction.

Size-resolved EF, number and mass
In Fig. 4, size-resolved EF PN for each bus class in the accelerating mode are shown, i.e. diesel buses with (Fig. 4a) and without (Fig. 4b) DPF and CNG buses (Fig. 4c).All classes show more or less a unimodal number size distribution.Diesel buses emit larger particles compared to CNG buses, peak diameter 70-90 nm and 10 nm, respectively, which is similar to results reported in Jayaratne et al. (2009) (80-90 nm and 10-12 nm, respectively).The lack of larger particles in the emissions from CNG-fuelled buses decreases the available surface area, hence favouring nucleation over adsorption/condensation of supersaturated vapours.This enhanced nucleation is one reason for the larger average particle number emissions for the tested CNG buses (Kumar et al., 2010).The mass size distribution shows that the diesel engines in the accelerating mode primarily emit particles with a diameter of ∼150 nm and that CNG buses exhibit on average a bimodal mass size distribution with one mode peaking at about 25 nm and another at ∼125 nm (Fig. 5).For the analysis of the average size-resolved EF PN/PM for buses without DPF (Figs. 4b and 5b), one bus (no.34) was excluded showing much higher size-resolved EF PN and with a peak size of ∼17 nm.For this bus the average size-resolved EF PM was bimodal with peak sizes of ∼30 nm and ∼190 nm.The reason for this discrepancy is not known but could be due to maintenance or malfunction of this particular bus.

Atmos
For the constant speed mode the characteristic bimodal number size distributions were obtained for the diesel buses with and without DPF, with one mode peaking at ∼10 nm (nucleation mode) and the other at ∼60 nm (soot mode/accumulation mode) (Fig. 6) (Maricq, 2007).The reason for the different average number size distributions between accelerating and constant speed mode may be more available surface area in the accelerating mode, hence favouring adsorption/condensation over nucleation.In acceleration from standstill the engine load is close to its maximum, and Jayaratne et al. (2009) also obtained a unimodal number size distribution for a diesel bus at 100 % load.

Comparison of EF part and EF gas (NO x , HC and CO)
The highest NO x values were obtained for the CNG buses compared to all the other Euro classes of diesel buses; however, the scatter was largest for the CNG buses as well (41 ± 26 g kg −1 ) (Fig. 7), which is in accordance with Ekström et al. (2005).Possible reasons for this variability may be vehicle maintenance and variations in the CNG composition (Shorter et al., 2005;Ayala et al., 2002).The EF for NO x ranged from 4 to 21 g km −1 depending on Euro class, which is in good agreement with reported values for HDVs and buses in the literature (Table 5).In comparison with the HBEFA 3.1 model, the measured values for EF NO x are on average lower for all the tested Euro classes but within the 95 % confidence interval for the Euro V with SCR and EEV buses.However, for some SCR-equipped buses and CNG buses higher EF NO x values were measured.One reason for some of the high values regarding SCR may be that it is critical that the exhaust temperature is high enough for the SCR to work properly.
In Fig. 8a there is a comparison of EF part and EF NO x ; both mass and number of particles show an anti-relationship with NO x , which is especially true when no DPF is installed.In a diesel engine there is a compromise between emissions of NO x and emissions of particles (Clark et al., 1999), as is demonstrated by the data in Fig. 8a.For the CNG-fuelled buses no such trend was observed.
Generally the emission of CO from a diesel engine is low as the combustion is carried out in an air-rich environment.This can be seen in the data for the tested buses, where the CO concentrations for many of the buses are below the de-  tection limit of the instrument (i.e.below 18 g (kg fuel) −1 ).However, for six of the buses CO concentrations were measured (3 times the std of the noise).In Fig. 8b the EF PM and EF CO are compared, and as is shown a positive relationship between EF PM and EF CO was observed.High CO concentration is an indication of incomplete combustion, hence favouring soot formation, i.e. high EF PM .Regarding number of particles there is also a positive relationship, however less profound (Fig. 8b) than the relationship between EF PM and EF CO .
The CO emissions are also influenced by DPF.The average EF CO for the diesel buses with DPF tested in this study, when assigning values below 6 (1 times the std of the noise) to 6 g (kg fuel) −1 , were 11 g (kg fuel) −1 (10 buses in total).For the buses without DPF the average EF CO was 14 g (kg fuel) −1 (18 buses in total); hence DPF is not only reducing particles but CO as well, as reported in Ayala et al. (2002) and Lanni et al. (2001).For the tested buses, DPF had no statistical significant effect on the amount of NO x emitted, which also is in agreement with results reported by Ayala et al. (2002).
Regarding total hydrocarbon (HC), emissions above the detection limit (14 g (kg fuel) −1 ) were not found for any of the buses in this study.Compared to the literature data shown in Table 5, values above the detection limit of our instrumentation were only reported for some CNG-fuelled buses.

Atmospheric implications and conclusions
The method of using a high time resolution particle instrument and CO 2 concentration as a tracer of the combustion source for determining EF PN and EF PM from individual vehicles for real-world dilution showed to be very successful regarding reproducibility, costs and number of vehicles studied.This method enabled measurements of not only particle number but also size, as well as mass.
Compressed natural gas buses are more advantageous regarding emissions of particle mass compared to diesel buses.However, in accelerating mode, generally CNG buses emit more particles by number compared to diesel-fuelled buses, and these particles are smaller (D p ∼10 nm compared to ∼80 nm) and presumably more volatile.The fact that CNG buses emit high number of particles in accelerating mode, e.g. at bus stops where many people may be standing waiting for buses, is an important aspect.However, the health impact of these particles versus diesel particles is still a matter of discussion.
This study shows that DPF markedly reduces emissions of particles both by mass and number as well as CO emissions also for real-world dilution.Reducing the number of soot mode particles does not cause a severe increase in nucleation mode particles as is the case for some of the tested CNG-fuelled vehicles without particle filter.
There was a large variation in NO x emissions from the tested SCR-equipped buses.This is most likely due to differences in engine and exhaust temperature, which influence the efficiency of the SCR to reduce NO x emissions.In particular this has implications for NO 2 population exposure in urban areas and is thus a health issue that needs to be investigated further.
Compared to other types of vehicles, the average EF PN for a diesel-fuelled bus without DPF is very similar to results obtained for a diesel passenger car without DPF (Hak et al., 2009) when looking at the number of particles emitted per kg fuel used (2.1 ± 1.0 × 10 15 kg −1 vs. 2.1 ± 0.3 × 10 15 kg −1 ).The mean EF PN for DPF-equipped diesel-fuelled buses were in the same order as an old petrol car (4.4 ± 3.5 × 10 14 kg −1 vs. 4.2 ± 3.0 × 10 14 kg −1 ) (Hak et al., 2009).However, when taking fuel consumption into consideration, there was a large difference.Diesel-fuelled buses without DPF are then emitting more particles per km −1 than a diesel passenger car without DPF, whereas DPF-equipped diesel buses are similar to a diesel passenger car without DPF (6.5 ± 3.2 × 10 14 and 1.4 ± 1.1 × 10 14 km −1 vs. 1.2 ± 0.2 × 10 14 km −1 ).On average the CNG-fuelled bus investigated in this study emitted a higher number of particles than a diesel passenger car both with respect to kg fuel burnt and per km driven.
In the data the typical trade-off trend between emission of NO x and particles (PN and PM) was observed, especially for vehicles without DPF, as well as a positive relationship between emissions of CO and PM/PN.
The data presented in this study demonstrate the variation in gas and particle emissions of the in-use fleet of a regional public bus service, where variations found between similar buses can be due to engine specifics, maintenance or malfunction.
Publications on behalf of the European Geosciences Union.5338 Å. M. Hallquist et al.: Particle and gaseous emissions from individual diesel and CNG buses

20Figure 2 .Fig. 2 .
Figure 2. Example of emission signals from three successive individual passages of the same bus in accelerating mode.Particle number (red line) and CO 2 concentration (black line).

Fig. 3 .
Fig. 3. EF PN (a) and EF PM (b) for all the buses studied divided into Euro class for the driving mode acceleration.Without DPF (white circles), with DPF (red circles), average of all represented Euro classes (dashed line), average of an individual represented Euro class (solid line).Crosses are EFs obtained by the HBEFA 3.1 model with DPF (red) and without (black).

Fig. 3 .
Fig. 3. EF PN (a) and EF PM (b) for all the buses studied divided into Euro class for the driving mode acceleration.Without DPF (white circles), with DPF (red circles), average of all represented Euro classes (dashed line), average of an individual represented Euro class (solid line).Crosses are EFs obtained by the HBEFA 3.1 model with DPF (red) and without (black).

Fig. 3 .
Fig. 3. EF PN (a) and EF PM (b) for all the buses studied divided into Euro class for the driving mode acceleration.Without DPF (white circles), with DPF (red circles), average of all represented Euro classes (dashed line), average of an individual represented Euro class (solid line).Crosses are EFs obtained by the HBEFA 3.1 model with DPF (red) and without (black).

Fig. 3 .
Fig. 3. EF PN (a) and EF PM (b) for all the buses studied divided into Euro class for the driving mode acceleration.Without DPF (white circles), with DPF (red circles), average of all represented Euro classes (dashed line), average of an individual represented Euro class (solid line).Crosses are EFs obtained by the HBEFA 3.1 model with DPF (red) and without (black).

Fig. 4 .
Fig. 4. Size-resolved average EF PN for diesel buses (Euro III-V) with DPF (a) and without DPF (b) and for CNG buses (c) for the driving mode acceleration.Solid lines represent averages and dashed lines the statistical 95 % confidence interval.For the data presented in graph (b) one bus (no.34) was excluded showing much higher size-resolved EF PN and with a peak size of ∼17 nm.

Fig. 5 .
Fig. 5. Size-resolved average EF PM for diesel buses (Euro III-V) with DPF (a) and without DPF (b) and for CNG buses (c) for the driving mode acceleration.Solid lines represent averages and dashed lines the statistical 95 % confidence interval.For the data presented in graph (b) one bus (no.34) was excluded showing a bimodal EF PM and with peak sizes of ∼30 nm and ∼190 nm.

Fig. 6 .Fig. 7 .
Fig. 6.Size-resolved average EF PN for diesel buses (Euro III-V) with DPF (a) and without DPF (b) for the driving mode constant speed mode.Solid lines represent averages and dashed lines the statistical 95 % confidence interval.Dotted lines represent averages for the accelerating mode.

Table 1 .
Technical data of the buses studied.
a Diesel = MK1 < 10 ppm S b SCR=selective catalytic reduction, EGR=exhaust gas recirculation, DPF=diesel particulate filter c Modified Euro III, now classified as Euro V d EEV=Enhanced Environmentally Friendly Vehicle e n.a = not available.

Table 2 .
HBEFA 3.1 Emission factors, fuel consumption (FC) and NO 2 to NO x ratios for Ubus Std > 15-18 t Urban Access Road/30/Stop + Go.NO x 10 14 (kg fuel) −1 g (kg fuel) −1 g (kg fuel) −1 g km −1 % * This NO 2 /NO x ratio has also been used in this study for Euro III buses with SCR and DPF.

Table 3 .
EF for particle number (EF PN ), mass (EF PM ) and gaseous compounds for all the buses studied in accelerating mode (acc) and constant speed mode (const).Stated errors are at the statistical 95 % confidence interval.
a In NO 2 equivalents b Modified Euro III, now classified as Euro V c Omitted when calculating average size distributions and total numbers d n.a = not available e Less than 8 g (kg fuel) −1 NO as NO 2 equivalents.

Table 4 .
Comparison of emission data for particle number and mass from present study with selected literature data.

5350, 2013 5346 Å. M. Hallquist et al.: Particle and gaseous emissions from individual diesel and CNG busesTable 5 .
Comparison of emission data for NO x , VOC and CO from present study with selected literature data.