Emission Measurement of Buses Fueled with Biodiesel Blends during On-Road Testing

: Increasing the biodiesel content of diesel fuels is encouraged because of its reduced carbon footprint. Pure rapeseed methyl ester (RME)and used cooking oil methyl ester (UCOME) are characterised by well-to-tank greenhouse gas (GHG) reductions of 54% and 88% compared to pure B0 petrodiesel, respectively. Captive ﬂeets such as public transport buses could beneﬁt from these GHG reductions by increasing the biodiesel content of their fuel because they have a consequent yearly fuel consumption. The aim of this paper is to compare on-road tailpipe emissions of a diesel bus when increasing the biodiesel concentration in the fuel. The tests were carried out on a standard city bus belonging to the Euro V EEV emission standard that was equipped with a portable emission measurement system measuring NO, NO 2 , PN, CO and CO 2 at the tailpipe. The bus followed the SORT which is representative of urban bus driving. The heavy urban on-road measurements indicated increased NOx emissions (24–26%), decreased PN emissions (43–45%) and slightly decreasing CO emissions for B30 RME and UCOME compared to B7. A measurement uncertainty analysis showed that the CO emissions were less reliable. Similar conclusions were drawn for the easy urban on-road bus emission measurements with smaller differences between B7 and B30 RME and UCOME.


Introduction
Biodiesel has a positive impact on CO 2 emissions with an estimated overall greenhouse gas (GHG) reduction of 30% to 90% compared to diesel depending on the type of biodiesel [1]. Provided that it is produced with effective processes and sustainable life-cycles (e.g., using domestic biomass) [2][3][4][5], biodiesel could contribute to the energy transition effort. In Western Europe, the diesel fuel used for vehicles and available at fuel stations may contain up to 7 vol.% of biodiesel. Increasing the volume concentration of biodiesel would reduce the carbon footprint of the fuel used by vehicles.
Captive fleets such as long-haul trucks and transit buses would in particular benefit from large scale GHG reductions when using a fuel with higher biodiesel concentration. These fleets are characterised by predictable and recurrent driving and refuelling patterns thus involving Rajaeifar et al. [17] highlighted the lack of reliable emissions measurements of urban buses fueled with biodiesel blends under real-world operating conditions in the literature. More specifically, a clear need for careful and complete experimental designs is identified taking into account covariate factors such as ambient conditions. In addition, Rajaeifar et al. [17] also noticed a lack of statistical analysis of the emission measurements. As an example, Merkisz et al. [21] compared emission measurements on a 18m bus fueled with pure diesel and B100 RME during standardised on-road test cycle (SORT) but did not include measurement uncertainties. This paper addresses these shortcomings by post-processing raw emission measurements with correction factors that take into account the ambient weather conditions. The innovative scientific contribution of this paper is to perform on-road emission measurements with a portable emission measurement system (PEMS) following representative conditions of urban bus driving. The real impact on emissions of biodiesel compared to diesel is thus quantified as opposed to the homologation engine dynamometer emission measurements.

Methods
This section first presents the fuel blends used during the experiments and then provide the details of the on-road tests.

Biodiesel Blends
The on-road test compares the emission performances of two B30 biofuels composed of 30 vol.% of either RME or UCOME with B7 fuel available at fuel stations. The latter is composed of 6.5 vol.% and 6.7 vol.% of Fatty Acid Methyl Esther (FAME) which may contain RME. The B7 fuel is used by public transportation companies for their diesel buses in Belgium. All fuels tested in this study are compliant with the EN590 and EN14214 norms. The sustainability of the renewable fuel supply chain is ensured by the international sustainability and carbon certification (ISCC) and the biomass biofuel sustainability scheme (2BS), both certifications defined by the European Commission. In addition, the biodiesel is compliant with the US Environmental Protection Agency [22]. RME is produced from vegetable oil by transesterification which currently provides the highest conversion efficiency at the lowest cost [17]. The Renewable Energy Directive II 2018/2001/EC promotes biofuels and biogas produced from used cooking oil for transport. Their energy content may be considered doubled in the target of reaching a 14% renewable energy share within the final consumption of energy in the transport sector by 2030. The Renewable Energy Directive II 2018/2001/EC defines an estimate of the GHG emission savings and total GHG emissions of biofuels for their cultivation, processing, transport and distribution and that are representative of the European Union consumption. Table 1 summarises these GHG emissions of rapeseed biodiesel and used cooking biodiesel of varying concentrations when produced with no net carbon emissions from land-use change [1]. A transition from B7 RME to B30 RME could already reduce the well-to-tank (WTT) GHG emissions by more than 10 g CO 2 eq /MJ which corresponds to a 14% reduction. Table 1. The GHG emissions for the cultivation, processing, transport and distribution of biofuels decrease as the volume concentration of RME increases. The reduction in WTT GHG emissions when shifting from pure fossil diesel (B0) to pure RME biodiesel (B100) or to pure UCOME biodiesel (B100) is 52% and 88%, respectively [1]. The composition and specifications of the B30 RME and B30 UCOME samples indicate only small differences between both fuels (Table 2). For both fuels, the carbon, hydrogen and oxygen content amounts to 83.5 molar %, 13.7 molar % and 2.84-2.85 molar %, respectively. The hydrogen and carbon atomic ratio were measured with the ASTM D5291 method and the oxygen atomic content was calculated based on the two previously mentioned atomic ratios. These values were measured on B30 RME and B30 UCOME samples that were produced in the same way at the same refinery as the samples used in the presented tests. Table 2. The specifications of the B30 RME and B30 UCOME samples indicate only small differences between both fuels. The composition and heat of combustion were measured on B30 RME and B30 UCOME samples that were produced in the same way at the same refinery as the samples used in the presented tests.

Method
Unit B30 RME B30 UCOME We performed on-road emission measurements on a 12 m diesel city bus. The tested bus is representative of the Western European bus fleets and has already covered nearly 400,000 km for TEC, one of the Belgian public transportation companies. The bus belongs to the Euro V EEV emission standard and has SCR and DPF aftertreatment systems. Technical specifications of the tested bus are detailed in Table 3. The empty weight of the vehicle amounts to 11,050 kg and the load present in the bus during the tests is estimated to 400 kg including the three people on the bus during the test, the emission measurement equipment, the secured driver cabin, and ticketing equipment. Emissions are measured following the SORT in order to compare in a reproducible way the emissions of the different fuels during typical driving patterns of transit buses in operation. These on-road test cycles have been defined by the UITP Bus Committee based on more than 10 million kilometres of operation data of transit buses and represent typical stop-and-go operation of a scheduled service bus as opposed to the engine tests required for vehicle certification [23]. The main aim of SORT is to have a reproducible comparison in fuel consumption of buses. These on-road test cycles have been defined since 2004 when real-operation on-road emission measurements with a PEMS were not yet required for in-service conformity tests. SORT is thus a precursor for on-road test cycles representative of buses' real operation.
The heavy urban SORT 1 and easy urban SORT 2 are selected for the measurement campaign given their higher impact on local air pollution than the suburban SORT 3. Urban operation covers more densely populated areas and is moreover characterised by a lower commercial speed. At average speeds lower than approximately 30 km/h, the fuel consumption of a bus increases as the commercial speed decreases [23]. The rated average speeds of base cycle SORT 1 is 12.1 km/h and SORT 2 is 18 km/h. SORT 1 and 2 are each composed of three simple trapezoidal speed profiles defined by an acceleration, a section at constant speed and a deceleration. It then ends with a stop, keeping the engine at idle (see Figure 1). More accurate specifications of the on-road test cycles are detailed in Table 4.     Although SORT is a standardised cycle and is not represented in full in real-world conditions, it is approaching real-world conditions since the test cycle is performed on the whole vehicle on the road instead of on an engine dynamometer as required for type approval. In addition, the test cycle is defined based on statistically generated data from several European transport companies (commercial speed, average time spent at stops, average distance between them, load, etc.) [23].
The load imposed by the SORT methodology for this particular bus in order to reach half load amounts to 3668 kg. A lack of space did not allow for adding in practice an additional load of 3.2 tons inside the bus given that the PEMS had to be accessible during the tests to verify regularly that no warning or errors were shown and took 10 seats in the back of the bus. However, the stiffer accelerations performed during the tests compared to the SORT cycle (see supra) can compensate the absence of load. Finally, given that all the tests were performed with an identical load, a comparison of the emissions of the different fuels can be done.

Portable Emission Measurement System
A PEMS by AVL is used for the on-road emission measurements. The PEMS consists of the following devices that are all connected to the AVL M.O.V.E. control system: • The AVL Gas PEMS is which is composed of NDIR and NDUV gas analysers for measuring carbon oxides (CO, CO 2 ) and nitrogen oxides (NO, NO 2 ), respectively. • The AVL PN PEMS is which consists of the modified particle detector Partector from Naneos working by DC. The PN counter is preceded by a VPR that is composed of a two stage dilution with an evaporation tube and catalytic stripper (300 • C) between the primary and secondary dilution [24]. The VPR is connected to the tailpipe with a short heated line (≤1 m) at 150 • C to avoid corrosion and the condensation of NO 2 and hydrocarbons. • The AVL EFM which gives a direct measurement of the instantaneous exhaust flow rate by using a pressure differential device (pitot tube). The EFM has a pipe size of four inches and measures flow rates between 30 kg/h and 2140 kg/h at 100 • C. • The Dearborn Protocol Adapter DPA 5 of DG Technologies which allows for record engine parameters such as the engine temperature and the vehicle and engine speeds via the OBD interface. • A Garmin GPS which provides vehicle localization, altitude, and ground speed.
• An ambient sensor measuring ambient temperature and relative humidity.
The tailpipe on the top of the bus is extended with the EFM to which a sampling probe is connected ( Figure 2). A heated line brings the sample exhaust gases through the roof trapdoor of the bus to the emission analysers inside the bus. The GPS and ambient sensor are installed on the roof of the bus. Post-processing calculations were performed with the AVL CONCERTO M.O.V.E. software according to the ISO16183 standard. A time alignment is firstly performed between the following signals: • the GPS and ECU velocity • the ECU throttle position and CO 2 measurement • the CO 2 measurement and exhaust mass flow measurement • the ECU throttle position and particle number measurement During the pre and post tests, zero and span adjustments of the gas analysers are performed with calibration gases. The measured concentrations are corrected for drift in the post processing according to the European Regulation EU R49 [25]. Drift-corrected raw exhaust gases are thereafter converted to wet concentrations. This results in comparable absolute measurements with the same content of water. The drift-corrected raw exhaust gases are hence multiplied with the dry-to-wet correction factor of the ISO16183 standard which is a function of the dry CO and CO 2 concentration of the exhaust gases and the absolute humidity (Appendix A). For the emission measurements, the dry-to-wet correction factor varied between 0.883 and 0.996. For NOx emissions only, an additional correction factor is applied to take into account the ambient conditions since a higher water content in the intake air reduces the formation of NOx during in-cylinder combustion [26]. The NOx correction factor for ambient conditions of the ISO16183 standard for compression-ignition engines is a function of the absolute humidity and the ambient temperature (Appendix B). For the emission measurements, the NOx correction factor for ambient conditions varied between 0.941 and 1.017. Instantaneous mass emissions are computed for each pollutant according to the EU legislation by multiplying the instantaneous concentration of the component in the exhaust gas with the instantaneous exhaust mass flow and a tabulated component specific factor. The latter one is the ratio between the density of the exhaust component and the density of the exhaust gas. The following equations detail the mass emission rate calculations:ṁ gas = X gas, DC CF dry/wetṁexh u gas (1) whereṁ gas is the mass flow rate of the measured gas (g/s), X gas, DC is the drift-corrected fraction of gas measured by the PEMS (ppm or %), CF dry/wet is the dry-to-wet correction factor (-),ṁ exh is the exhaust mass flow rate (kg/s), u gas is the component specific factor (-) and CF NOx is the NOx correction factor for ambient conditions (-). The distance-specific emission E gkm of one SORT cycle is computed as follows: where fṁ is the sampling frequency of the mass flow rate and d ecu is the distance of a SORT repetition computed from the ECU vehicle velocity.
For particle number emissions, no correction factor is applied in the post-processing because the amount of particles present in the exhaust gases is not influenced by the varying concentration of gases including water. The VPR and heating lines ensure that only non-volatile components are entering the PN measurement device.

Uncertainty Quantification
The physical quantity E gkm of the measured gas emissions involves combined uncertainties because it is computed from the following 6 (or 7 for NOx emissions) measured quantities: • X gas, DC , drift-corrected fraction of gas measured by the PEMS, •ṁ exh , the exhaust mass flow rate, • X CO, DC and X CO 2 , DC , the drift-corrected fraction of CO and CO 2 measured by the PEMS, • H a , the absolute humidity, • d ecu , the trip distance computed from the ECU vehicle speed, • T amb , the ambient temperature (for NOx emissions only).
Considering that these 6 (or 7 for NOx emissions) measurements are uncorrelated and that the second order derivative factor is negligible, the global uncertainty of the physical quantity E gkm = f(W 1 ,...,W 6/7 ) is expressed with Equation (4) [27]. A normal distribution is considered for the rate of emission per kilometre E gkm of one SORT cycle. The maximum uncertainty of E gkm with a 95% probability is thus computed as follows: Only the accuracy is considered for the uncertainties of the emission measurements because the linearity, repeatability and span and zero drift of the gas analysers are much smaller [28][29][30]. Table 5 lists the accuracy of the measurement devices. Table 5. The accuracy of the gas analysers and the EFM is small compared to the experimental variability and does not exceed 2.5% of the analysers' measurement range. The linearity, repeatability and span and zero drift of the gas analysers is neglected in this uncertainty quantification [28][29][30]. The RDE Commission Regulation 2016/427 imposes a maximum deviation of 4% for the total trip distance calculated from the ECU vehicle speed between the measurement instrument and the distance as calculated with a topographic map. Hence, a conservative and worst case accuracy of 4% is taken into account for the trip distance computed from the ECU vehicle speed. An accuracy of ±0.4 • C and 4% of the measurement is considered for the ambient temperature measurement and the absolute humidity, respectively [31].

Accuracy
The maximum uncertainty with 95% probability computed for the SORT 1 and 2 test runs with B7, B30 RME, and B30 UCOME are 10.7 g/km for CO 2 and 0.14 g/km for NOx. This measurement uncertainty does not take into account the variance of the measured emissions which will be highlighted on the graphical representation.
For the particle number emissions, the measurement uncertainty of the particle detector Partector is predominant indicating an accuracy of ±1000 #/cm 3 or ±30% of the measurement range (0-10 6 #/cm 3 ), whichever is greater. As simplification, we consider a conservative accuracy of ±30% of the particle number emission rate per kilometre.

Quality Assurance: Trip Validity SORT
The repetitions of the experimental SORT cycles are compared with the SORT cycle guideline by firstly verifying the speed profile per unit of distance (Figure 3). The velocity profiles of the repetitions slightly deviate from the theoretical SORT guideline, but the different test runs show a good reproducibility. The more severe accelerations of the experimental test runs are responsible for shorter acceleration periods and earlier idling times than the SORT guideline because the target velocity of the speed trapeze is reached earlier.   Lastly, a comparison of the emission measurements of the different test runs shows that the differences between test runs regarding driving patterns does not largely affect the emission measurements ( Figure 5). During accelerations, large increases in NOx emissions are observed and idling periods are characterised by the same stabilised emissions for different test runs. The cumulative NOx emissions show the good reproducibility of the SORT 1 test runs.
The velocity profiles and dynamics of the driving patterns as well as the emission measurements of the different test runs show good reproducibility, validating a comparative analysis of the different fuels.

Results and Discussion
During the heavy urban SORT 1 cycle, increased nitrogen oxide emissions are measured when the bus is fueled with B30 RME and B30 UCOME compared to B7. In Figure 6, each point represents the NOx distance specific emission measured during one SORT 1 cycle repetition. A swarm plot is selected for representing the measurements in order to show the spread of the repetitions and the overlapping points clearly. The increase in NOx emissions is due to the higher oxygen content of biodiesel compared to petrodiesel, thus promoting the combustion and heat release rate. NOx formation increases exponentially with the combustion temperature as observed with the Zeldovich mechanism. The mean of the NOx emissions per SORT 1 run increases approximately by 25% for B30 RME and B30 UCOME compared to B7. 10 Figure 6. The NOx emission rate per kilometre of each SORT 1 repetition shows higher NOx emissions when the bus is fueled with B30 RME and B30 UCOME compared to B7. The maximum measurement uncertainty with 95% probability amounts to 0.14 g/km . Figure 7 shows the cumulative NOx emissions of each SORT 1 repetition with the three different fuels. This representation indicates that the different SORT test runs are consistent confirming the quality assurance of the SORT repetitions discussed in Section 2.2.4. All SORT 1 repetitions indicate steep increasing NOx emissions during accelerations (@0 s, 42 s, 93 s) and close to constant NOx cumulative emissions during idling (@22-42 s, 73-93 s, 120-140 s). In addition, Figure 7 shows a progressive increase in cumulative NOx emissions for B30 RME and B30 UCOME compared to B7. The increase in NOx emissions for B30 fuels is thus not caused by one particular event such as an aggressive acceleration of the SORT cycle. There is an increase of 1.33 g (±0.8 g) of the mean cumulative NOx emissions at the end of SORT 1 when switching from B7 to B30 RME.

SORT1
B30 RME B30 UCOME B7 Figure 7. The cumulative NOx emissions along the SORT 1 test runs indicate an increase in NOx emissions for B30 RME and B30 UCOME compared to B7.
The NOx emissions measured during the easy urban SORT 2 cycle follow the same trend as the SORT 1 test runs with higher NOx emissions for B30 RME and UCOME than for B7 ( Figure 8). However, the increase in NOx emissions of the B30 RME and UCOME is less important with average increases of 10 to 14% compared to B7. In addition, the NOx emission rate per kilometre measured during SORT 2 test runs is characterised by a large variance and is lower than during the SORT 1 tests. This is due to the lower accelerations and higher average speed of the SORT 2 cycle. The more aggressive driving dynamics of SORT 1 result in a higher fuel consumption and a higher NOx emission rate per kilometre.  Figure 8. The easy urban SORT 2 (less aggressive dynamics) test runs indicate higher NOx emissions for B30 RME and B30 UCOME compared to B7 with an average increase varying between 10% and 14%. Figure 9 highlights the same conclusions for the cumulative NOx emissions of the SORT 2 repetitions as those discussed for SORT 1. The cumulative NOx emissions at the end of the SORT 2 cycle are higher than those of the SORT 1 cycle due to the longer distance covered by the SORT 2 cycle. However, relatively, the increase in cumulative NOx emissions is lower for SORT 2 than for SORT 1.

SORT2
B30 RME B30 UCOME B7 Figure 9. The cumulative NOx emissions along the SORT 2 test runs indicate, alike for SORT 1, an increase in NOx emissions for B30 RME and B30 UCOME compared to B7.
Higher tailpipe NOx emissions of fuels with higher biodiesel concentrations are attributed to both increased NOx engine out emissions and a lower NOx reduction efficiency [32]. The NOx reduction performance depends on the NO 2 /NOx ratio at the SCR inlet with an optimal value of 0.5 [32]. When fueled with B100 the NO 2 /NOx ratio decreases to 0.1-0.2 due to NO 2 deficiency.
When the NO 2 /NOx ratio drops below 0.4, the NOx reduction efficiency drops sharply. Slightly lower exhaust-gas temperatures are measured for B100 compared to B0 resulting in lower oxidative power of the Diesel Oxidation Catalyst (DOC). Consequently, less NO will be transformed to NO 2 which reduces the NO 2 /NOx ratio at the inlet of the SCR for B100 compared to B0. Mitigation strategies for NOx increases include among others to add cetane improver additive to the biodiesel blend, change injection timing or add a lean NOx catalyst [33].
Concerning particle number emissions, a decrease is measured during the heavy urban SORT 1 cycle when the bus is fueled with B30 RME and B30 UCOME compared to B7. In Figure 10, each point represents the PN emission rate per kilometre measured during one SORT 1 cycle repetition. The decrease in particle number emissions is due to the higher oxygen content of biodiesel compared to petrodiesel. In addition, the better combustion and higher heat release rate of the combustion of B30 RME and B30 UCOME is also responsible for the reduction in particulate matter emissions. The mean of the PN emissions per SORT 1 run decreases approximately by 44% for B30 RME and B30 UCOME compared to B7.
The cumulative PN emissions of each SORT 1 repetition confirm the consistency of the repetitions and indicate a progressive and pronounced decrease in PN emissions when fueled with B30 RME and B30 UCOME compared to B7 ( Figure 11). Considerably steep increases in PN emissions are measured during accelerations (@0 s, 42 s, 93 s) for all SORT 1 repetitions. During constant speed, deceleration and idling, the cumulative PN emissions remain relatively constant. There is an increase of 7.7 × 10 11 particles (±2.6 × 10 11 particles) of the mean cumulative PN emissions at the end of SORT 1 when switching from B7 to B30 RME.  Figure 10. The PN emission rate per kilometre of each SORT 1 repetition shows lower PN emissions when the bus is fueled with B30 RME and B30 UCOME compared to B7. SORT1 B7 B30 UCOME B30 RME Figure 11. The cumulative PN emissions along the SORT 1 test runs indicate a progressive decrease in PN emissions for B30 RME and B30 UCOME compared to B7.
Concerning the PN emissions measured during the easy urban SORT 2 repetitions, the difference between the B30 RME and B30 UCOME fuels is analogous to those discussed for SORT 1 (Figures 12 and 13). In Figure 12, the average decrease in PN emission rate per kilometre for B30 RME and B30 UCOME varies between 24 and 36% during SORT 2 repetitions. Alike for NOx emissions, a less pronounced difference in PN emissions between fuels is observed during SORT 2 repetitions compared to SORT 1 repetitions. A possible explanation for the difference in PN emssions between B30 RME and B30 UCOME during SORT 2 is the more efficient combustion occurring during SORT 2 because higher speeds are reached combined with the better quality of UCOME compared to RME. A pretreatment is indeed performed on used cooking oil before transesterficiation to eliminate their content of impurities such as Free Fatty Acid (FFA), water and solids resulting from their first use. To eliminate water content, used cooking oil is heated to above 100 • C or vacuum distillation is performed at a 0.05 bar pressure [34]. Suspended solids, phospholipids and other impurities are to be washed away with hot water or removed by centrifugation and paper filtration [34]. SORT2 B7 B30 UCOME B30 RME Figure 13. The cumulative PN emissions along the SORT 2 test runs indicate, alike for SORT 1, a decrease in PN emissions for B30 RME and B30 UCOME compared to B7.
Higher DPF regeneration rates and a lower balance point temperature are observed when using biodiesel blends [35]. The balance point is indicative of the performance of a catalysed DPF and is defined as the temperature at which the cDPF is oxidizing soot at the same rate that is is being deposited. The difference in DPF performance of soot emissions produced from biodiesel as opposed to diesel engine combustion is due to the higher reactivity in oxygen of biodiesel soot which results on its turn from a more disordered soot structure and higher levels of oxygen [35].
Regarding carbon monoxide emissions, there is no considerable difference measured between B30 RME, B30 UCOME and B7. In Figure 14, the CO emission rate per kilometre of SORT 1 repetitions are spread out on a similar range between 0.86 and 6.62 g/km. The high variance in CO emissions for the three fuels during SORT 1 repetitions is also highlighted by the cumulative CO emissions (Figure 15). CO concentrations on buses are low and close to the lower bound of the CO measurement range. As illustration, the AVL Gas PEMS used in the measurements has an absolute measurement accuracy of 30 ppm for CO which corresponds to the lower CO emission measured on the bus. The average CO emissions for the SORT 1 repetitions of B30 RME and UCOME tend to indicate a reduction compared to B7. Due to the increased oxygen content in the B30 fuels and thus better combustion, less unburned gases are expected.
B30 UCOME B30 RME Figure 15. A large spread in cumulative CO emissions is measured among the different SORT repetitions without regard for the used fuel.
Concerning carbon dioxides emitted at the tailpipe, a similar range of emission rate per kilometre is measured on the bus when fueled with B30 RME, B30 UCOME and B7. During SORT 1 repetitions, CO 2 emissions varied between 1193 g/km and 1311 g/km ( Figure 16).
A comparison of the relative emission differences between the SORT performed in the framework of this paper and the trends of the HD engine and chassis dynamometer tests reviewed in the introduction indicate more pronounced emissions increases or decreases for the on-road SORT between the biodiesel and conventional diesel fuels. Several factors could influence the change in the emission measurements such as the performed test, the experimental variability and tested engines. A possible cause for the difference in importance of the trends between the SORT tests and the engine and chassis dynamometer is the lack of representativity of real use of engine and chassis dynamometer tests and optimisation of the engine/vehicle for engine/chassis dynamometer tests.

1193.3
1310 B30 RME B30 UCOME SORT1 Figure 16. No considerable difference in CO2 tailpipe emissions is measured during SORT 1 when the bus is fueled with B30 RME, B30 UCOME and B7. The maximum measurement uncertainty with 95% probability amounts to 10.7 g/km .

Conclusions
This work compares the tailpipe emissions measured on a public transportation bus and heavy-duty engine fueled with increased concentrations of RME and UCOME. Captive fleets such as public transport buses could benefit from a lower carbon footprint by using fuels with higher biodiesel content such as B30 (containing 30% in volume of biodiesel). This change not only requires sustainable biodiesel; it also requires to know the impact of B30 fuels on local tailpipe emissions in operation. Hence, this study assesses precisely, based on representative on-road emissions tests and taking into account the measurement uncertainty, what the impact is of fuels with higher concentrations of biodiesel on local tailpipe emissions of buses.
During the on-road heavy urban bus driving, an increase of 24-26% NOx emissions is measured for B30 RME and B30 UCOME compared to B7. A smaller increase in NOx emissions with larger variance is measured during the easy urban bus driving that is characterised by a higher average speed. particle number emissions are decreased by 43-45% during the heavy urban bus driving when fueled with B30 RME and B30 UCOME compared to B7. The CO emissions measured during the on-road emissions test are very low and not significant when considering the measurement uncertainty. Equivalent CO 2 emissions are measured on the road for B30 RME, B30 UCOME and B7 as they are characterised by a large variance in the cycle measurements. These on-road emissions measurements are compared with engine dynamometer emissions measurements on heavy-duty engines from the literature indicating less important trends during engine dynamometer emission measurements from the literature compared to SORT. Finally, the effect of aftertreatment systems on emissions from biodiesel are investigated and explained.
Future work should study the calibration of the SCR aftertreatment for higher NOx emissions as reported for the B30 fuels in this paper during representative on-road tests. This would assure that the benefit in carbon footprint of fuels with higher concentrations of biodiesel would not be accompanied by an increase in NOx tailpipe pollutants.

Conflicts of Interest:
The authors declare no conflict of interest.