Atmospheric Impact of passenger car NO X emissions on urban NO 2 pollution – Scenario analysis for 8 European cities

necessary to decrease urban air pollution. We regard our model results as robust albeit subject to uncertainty resulting from the application of a generic street layout. With small modi ﬁ cations in the input parameters, our model could be used to assess the impact of NO X emissions from road transport on NO 2 air pollution in any European city.


Introduction
Residents of many larger European cities are still exposed to NO 2 concentrations that often exceed the established air quality standards (EEA, 2015). Most exceedances occur in city centres, mainly caused by traffic-related NO X (nitrogen oxides) emissions originating from diesel cars. Policy makers expected the NO X emissions of diesel cars to decrease considerably when introducing more stringent Euro 5 and Euro 6a/b emission limits in 2009(EC, 2012, 2008a. This expectation, however, has not materialized. Tests with Portable Emissions Measurement Systems (PEMS) suggest that Euro 5 and 6a/b diesels, just as Euro 3 and 4 diesels before, emit on the road several times more NO X than permitted by the applicable limit (Weiss et al., 2012;Franco et al., 2014;Kadijk et al., 2015). Also tests carried out by the national type approval authorities in the aftermath of the dieselgate scandal have shown that on the road Euro 5 and 6a/b cars emit on average five times more NO X than their respective limits of 180 and 80 mg/km (Degraeuwe and Weiss, 2017;Transport and Environment, 2016). In view of the persistent air quality problems, the so-called Real-Driving Emissions (RDE) on-road test procedure with PEMS was introduced in the Euro 6 emission legislation (EC, 2017(EC, , 2016. Euro 6c introduces RDE for monitoring only without quantitative requirements for all new vehicles form 1/9/2018. Euro 6d-TEMP introduces binding RDE limits with a temporary conformity factor (CF) of 2.1. This means that NO X emissions measured on the road have to be below 80 × 2.1 = 168 mg/ km. From 1/9/2017 this limit applies to new types and from 1/9/2019 to all new vehicles. Regulation EU2017/1151 (EC, 2017) specifies that the WLTC (Worldwide harmonized Light vehicles Test Cycle) replaces the NEDC for emission testing on the test bench from 1/9/2017 for new types and from 1/9/2018 for all new vehicles. With Euro 6d the CF will be lowered to 1 plus measurement uncertainty. Initially the uncertainty is set at 0.5 but will be revised annually. This means the NO x limit will be 120 g/km. Euro 6d will be introduced on 1/1/2020 for new types and in 1/1/2021 for new vehicles. For on road NO x emissions three subclasses of Euro 6 are important: Euro 6a-c, Euro 6d-TEMP and Euro 6d.
In spite of these provisions, it yet remains unclear whether the implemented conformity factors will effectively mitigate urban NO 2 (nitrogen dioxide) pollution (EEA, 2013).  found in a case study for Antwerp (Belgium) that exceedances of the NO 2 air quality standard can be reduced substantially if diesel cars emitted on the road less than 80 mg NO X /km. However, a more comprehensive assessment that takes a continental perspective and analyses the impact of the RDE conformity factors on the air quality in major European cities is still missing. Here, we address this knowledge gap by modelling urban NO 2 pollution for various NO X emission scenarios. We pursue two objectives: • analysing the reduction in urban NO 2 pollution resulting from a decrease in the distance-specific NO X emissions of Euro 6 diesel cars; • assessing whether a dedicated NO 2 emissions limit, applied next to the existing NO X limit, could help reducing urban NO 2 pollution more effectively.
To this end, we generalise the air quality model developed by  and apply it in a scenario analysis to 8 European cities. The outcome can assist the annual revision of RDE conformity factors and provide scientists with a simplified, yet accurate, tool to model NO 2 pollution for any city in Europe. The article continues in Section 2 with an explanation of methods and data sources. We present and discuss the results in Sections 3 and 4. The article finishes with conclusions for scientists and policy makers in Section 5.

General overview
Our modelling of urban NO 2 pollution is based on the method applied by . We briefly explain this method in Section 2.2 and assess in a comparative case study for Antwerp (Belgium) and Milan (Italy) the sensitivity of the calculated NO 2 concentrations to key input parameters (Section 2.3). This assessment provides scope and justification for the introduction of model simplifications that allow us to apply the method to 8 European cities. Modelling of NO 2 pollution at the scale of a city requires large amounts of data; compiling and processing these within this research for a larger number of cities is unfeasible. We explain the simplified modelling approach applied here in Section 2.4.

Overview of the NO 2 pollution model applied to Antwerp
The NO 2 air pollution modelling of  uses the NO X emissions given for the year 2009 by the latest version of the TNO-MACC-II emissions inventory (Kuenen et al., 2014). These 2009 NO X emissions constitute the baseline emissions scenario for subsequent pollution modelling. In a second step, nine emission scenarios for road vehicles are defined, assuming a situation where virtually the whole light-duty and heavy-duty vehicle fleet consists of Euro 6/VI-compliant vehicles as it may be the case around the year 2030. The vehicle emission scenarios include (i) two synthetic worst-case scenarios that assume both high NO X emissions and high NO 2 fractions, (ii) a best-case scenario that assumes a complete fleet shift to gasoline cars and (iii) six scenarios with different combinations of NO X emission factors and NO 2 fractions (see 2nd column of Table 3). The emission factors cover the whole spectrum of Euro 6, from a to d. The assumed emission factors are based on the following information: (i) for light-duty diesel vehicles, on-road emission measurements as published by Franco et al., 2014); May et al., 2014;Vlachos et al., 2015;Weiss et al., 2012 andthe RDE legislation (EC, 2017). (ii) for heavy-duty vehicles, on-road compliance with the Euro VI limit was assumed, as suggested by the measurements of Vermeulen et al. (2012).
The gradual implementation of Euro 6, from Euro 6a to d, is not modelled. The scenarios have to be considered as an asymptotical situation in which the major part of the fleet complies with Euro 6d. Hence also the year 2030 is indicative. Another study for Germany (Toenges-Schuller et al., 2016) takes the fleet evolution into account which allows to estimate when the NO 2 limit will be reached.
The future vehicle activity and the share of diesel cars in the vehicle fleet are retrieved from the TREMOVE model (Breemersch et al., 2010). By combining data on the distance-specific NO X emissions of light-and heavy-duty vehicles, vehicle activity, the share of diesel cars in the vehicle fleet, the overall NO X emissions from road transport for all European countries are modelled. The resulting emission data are then used as input for the chemical transport model LOTOS-EUROS (Schaap et al., 2008). A run at 28 × 28 km resolution for the entire European continent was done for the baseline and the nine NO X emission scenarios. This run is then complemented by more detailed modelling at a 7 × 7 km resolution for the Benelux region that is used to establish the background NO 2 pollution level of Antwerp. To model the NO 2 pollution at high resolution in the city of Antwerp the IFDM (Immission Frequency Distribution Model) is used. IFDM is a bi-Gaussian plume model developed by Lefebvre et al. (2011). The model uses both the urban background concentration from LOTOS-EUROS (Schaap et al., 2008) and the NO X emissions at street level to produce a high resolution concentration map for the city. For detailed street canyon calculations, IFDM incorporates the OSPM Model (Berkowicz et al., 1997;Ketzel et al., 2012). The latter uses the rooftop concentrations determined by IFDM, street canyon geometry, and the NO X emissions in the canyon to calculate both hourly and annual average NO 2 concentrations in the specific street canyon under consideration.

Case study on the sensitivity of modelled NO 2 concentrations
The method described above requires detailed city-specific input on, e.g., the NO 2 background concentration, wind speed, temperature, building heights, street widths, fleet composition, and traffic intensity. To apply this model to other cities requires either collecting the necessary data for each individual case or applying generic assumptions. The latter approach reduces the demand for data and resources but inevitably introduces uncertainty into the NO 2 air pollution modelling. We consider it unfeasible, to collect within the scope of our research data, e.g., on building heights, street widths, and traffic intensities for a large number of European cities. We therefore assess via a case study for the cities of Antwerp and Milan the scope to simplify modelling parameters and based on this develop and apply generic assumptions to other cities. To this end, the following parameters are considered to influence the modelled NO 2 concentrations at the street canyon level: (i) hourly wind speed, (ii) hourly temperature, (iii) latitude dependent solar irradiation, (iv) the composition of the vehicle fleet and (v) background NO 2 concentration. We chose Milan for this case study because the conditions in this city differ considerably from those in Antwerp. In fact, the difference in the conditions between both cities appears to cover large parts of the wide range of climatic conditions and fleet compositions present in Europe: • The average annual wind speeds in Antwerp and Milan are 6.5 km/h and 2.4 km/h, respectively, covering approximately the range of wind speeds observed in Europe. The North Sea coast is one of the windiest places in Europe (only Scotland is windier); the Po Valley has approximately the lowest wind speeds in Europe (DTU Wind Energy, 2015).
• The share of diesel cars in the passenger car fleet in Antwerp is very high (85%) while it is near the European average in Milan (64%). We conduct a total of 7 model runs; 5 runs altering the 5 parameters individually, one run with the 3 weather parameters together (wind, temperature, latitude), and a final simulation run with the Antwerp street configuration and the climatic conditions, vehicle fleet composition, and NO 2 background concentration of Milan. As the results of our sensitivity study suggest, model simplifications are feasible (see Section 3.1). The generic assumptions made to estimate the NO 2 air pollution in 8 European cities are explained next.

Simplified modelling of NO 2 pollution in 8 European cities
We determine the effect of diesel car NO X emissions and their NO 2 fractions on the NO 2 air pollution in the following 8 cities: Antwerp, Athens, Barcelona, Krakow, London, Milano, Paris, and Stockholm. The chosen cities (i) comprise major metropolitan areas with a large population, (ii) resemble a wide range of geographic and climatic conditions across Europe, and (iii) are frequently exposed to elevated NO 2 pollution levels (EEA, 2013).
We base our modelling, on data for the local vehicle fleet composition as obtained from the TREMOVE model (Breemersch et al., 2010). The local weather conditions (i.e., wind and temperature) are obtained from the LOTOS-EUROS simulations (Schaap et al., 2008). Likewise, the NO 2 background concentration for each city and emission scenario is obtained from the LOTOS-EUROS simulations (see Section 2.2). Besides the Europe-wide simulation with a 28 × 28 km resolution, we conduct for three regions (i.e., Benelux, Po-Valley and South Poland) simulations at a 7 × 7 km resolution. These three regions are chosen because they constitute hot spots of high NO 2 air pollution. For the cities located in one of these three areas the 7 × 7 km runs are used, for the others the 28 × 28 km runs. We calculate the NO 2 concentration in a generic street canyon for each of the 8 cities, by applying the IFDM (Lefebvre et al., 2011) and OSPM (Berkowicz et al., 1997;Ketzel et al., 2012) models.
Following the outcome of our sensitivity study (see Section 3.1), we will use the street grid of Antwerp combined with the local boundary conditions (Table 1) for all 8 cities. This simplification is justified because a street grid represents a generic collection of street canyons with different orientations, building heights and traffic intensities rather than the precise street map of a city. The assumption of a generic street grid allows controlling for the effect of city geometry on the dispersion of NO 2 pollution for the range of assumed emission scenarios under a location-specific set of boundary conditions. Yet, the simplification also comes with a loss of accuracy which we discuss in Section 4.
For each city, we model at high-resolution the NO 2 pollution in a typical street canyon with a width of 50 m and a building height of 18 m. The generic canyon has a two-way single carriageway with 2 drive lanes per direction. The orientation is east-west. We assume a Table 1 Annual average NO 2 background concentrations and meteorological variables for 8 European cities; concentrations and meteorological data obtained from the LOTOS-EUROS model (Schaap et al., 2008) for the reference year 2009; percentage of kilometres driven with diesel cars from TREMOVE (Breemersch et al., 2010   peak-hour traffic of 870 vehicles per hour per lane, which is close to the maximum hourly capacity of a lane of around 1000 vehicles (UKHA, 1999). It is representative of a busy access road to a city centre. For this canyon, we likewise apply the ten emission scenarios described in paragraph 2.2. We compare the modelled background NO 2 and O 3 concentrations with the measured ones for all 8 cities. For the baseline year 2009, we establish a correlation between the modelled and measured daily average concentrations. Based on this correlation, we correct our modelled NO 2 concentrations of the various emission scenarios. We continue in the next section by presenting the results of our sensitivity study and the findings for the NO 2 background and street canyon concentrations in the 8 selected cities.

Resultssensitivity study for Antwerp and Milan
By applying the set of boundary conditions (i.e., climate, NO 2 background concentration, and composition of the vehicle fleet) from Milan to the street grid of Antwerp, we find that the annual average NO 2 concentration in Milan exceeds the one in Antwerp by 23 μg/m3 ( Table 2). The low wind speed and the high NO 2 background concentration each increase the NO 2 concentration by 15 μg/m 3 and 12 μg/ m 3 , respectively. In fact the frequently stagnant air in Milan causes both higher NO 2 background concentrations and higher NO 2 concentrations in street canyons compared to Antwerp. By contrast, the higher average temperature in Milan (14.7°C) compared to Antwerp (10.8°C) has virtually no effect on the average NO 2 concentrations. Likewise, the lower share of 63% versus 85% of passenger vehicle kilometres driven by Diesel cars in Milan compared to Antwerp only results in a 1 μg/m 3 reduction in ambient NO 2 concentrations. This seems a small effect but it is worth noting that the background is heavily influenced by the diesel share. Likewise, The 99.79 th percentile of hourly NO 2 concentrations in both cities, which is a measure for the exceedance of the hourly limit of 200 μg/m 3 , show the same behaviour as the annual average NO 2 concentrations in both cities ( Table 2). The annual average ozone concentration in Milan is 5 μg/m 3 higher than in Antwerp. Two compensating effects cause this: The comparatively low wind speed in Milan causes higher NO concentrations that, in first instance, reduce the ozone concentrations by 8 μg/m 3 . However, this effect is compensated by a background ozone concentration that is higher in Milan than in Antwerp, which leads to an increase in the ozone concentration of 13 μg/m 3 . Based on these findings, we conclude the most influential parameters are the wind speed, the background concentration and the vehicle fleet.

Impact of scenarios on the urban NO 2 background concentration
The background urban NO 2 concentration decreases with a reduction in NO X emissions per-km from diesel cars. The example of the Po Valley suggests a decrease of up to 15 μg/m 3 if the NO X emissions of diesel cars decreased from 550 to 100 mg NO X /km (Fig. 1). This impact is not limited to big cities only but extends over large areas (see also . The decrease in NO X emissions also causes a decrease of the 93 rd percentile of the O 3 maximum daily 8-h mean ozone concentration. The limit value of 120 μg/m 3 is exceeded mostly around the Mediterranean (EEA, 2015). E.g. scenario 5 (shift to gasoline) gives an improvement of about 20 μg/m 3 in the Po Valley where values above 140 μg/m 3 are common.
Likewise, an increase in the NO 2 fraction of the NO X emissions of diesel cars can increase the regional NO 2 background concentrations. A comparison between Scenarios 7 and 10 (NO x emissions of 550 mg/km but different NO 2 fractions of 20% and 70%, respectively) suggest the annual average NO 2 concentration can increase by up to 8 μg/m 3 when the NO 2 fraction in the NO X emissions is increased from 20 to 70% (Fig. 2). Around big urban areas (Milan and Turin) a decrease in the NO 2 fraction also causes the ozone concentrations (annual average and 93 rd percentile) to decrease. Contrary to the effect of the NO X emissions, the effect of the NO 2 fraction is more localized around the big emission sources.

Urban NO 2 concentrations at street canyon level
The NO 2 concentrations in the generic street canyon vary substantially between the 8 cities for the various scenarios (Table 3). In the baseline scenario, Milan (62 μg/m 3 ), London (57 μg/m 3 ), and Barcelona (56 μg/m 3 ) show the highest annual average NO 2 concentrations. These cities also show the highest background NO 2 concentrations ( Table 4). The behaviour of the NO 2 concentrations (annual average and the 99.79 th percentile as indicator for compliance with the hourly limit of 200 μg/m 3 ) as function of the assumed scenarios on NO X and NO 2 fractions is similar for all 8 cities. For cities where diesel cars contribute a large part to the urban NO 2 concentrations (e.g., Antwerp  and Paris), reductions in the NO X emissions of diesel cars translate directly into substantial reductions in NO 2 concentrations. A reduction in the NO X emissions of diesel cars to Euro 6 levels (80 mg/km) would bring NO 2 concentrations in all 8 cities below the European air quality standard (Table 3). However, if the on-road NO X emissions of diesel cars remain at the current levels (compare baseline scenario with Scenarios 7 to 10), only small reductions of around 10% in the ambient NO 2 concentrations can be expected at best. Fig. 3 (left) depicting the specific situation in a street canyon in Milan suggests that a reduction in the NO X emissions yields a considerable decrease in the annual average NO 2 concentration while changes in the NO 2 fraction has little impact on the NO 2 concentration. However, the number of exceedances of the hourly limit of 200 μg/m 3 , modelled here as the 99.79 th percentile of the NO 2 concentrations, increase with an increasing NO 2 fraction in the NO X emissions Fig. 3 (right). It is even possible that increasing NO 2 fractions increase the NO 2 concentration in the street canyon, despite overall decreasing NO X emission levels (compare Scenario 1 and 10). This finding confirms the observation of  for Antwerp and applies to some extent to all 8 cities covered by our analysis. The effects of a change in NO X emissions and their NO 2 fractions on the NO 2 concentration decreases with a decreasing share of diesel cars in the vehicle fleet (compare the results in Table 3 for Athens (9% share of diesel cars in the passenger car fleet) versus Antwerp (85% diesel cars in the passenger car fleet)). The observed reductions in NO 2 concentrations in the case of Athens mainly result from lower NO 2 background concentrations and decreased NO X emissions from heavyduty vehicles. Likewise, a high contribution of the NO 2 background to the NO 2 concentration in the city street canyon (as is the case for Krakow) decreases the importance of the NO X emissions from diesel cars for urban air quality.

Strengths and limitations of our research
This article assesses the effect of NO X emissions of Euro 6 Diesel cars on the NO 2 pollution in 8 European cities. Our modelling is based on the method developed by  and applies cityspecific conditions (such as wind speed, temperature, and vehicle fleet composition) to a generic street grid. Our modelling captures the wide range of scenarios for both NO X emissions and NO 2 fractions, thereby providing policy makers with clear indications about the impact of vehicle emissions on NO 2 air quality in the time horizon until 2030. The assumed model simplifications reduce the demand for data while still allowing us to account for relevant phenomena of pollution transport and atmospheric chemistry at European scale down to the level of the street canyon. With small modifications in the input parameters, our model can be used to assess the impact of NO X emissions from road transport on NO 2 air pollution in any European city.
Still, our research is also subject to limitations. The assumption of a generic street grid and NO 2 background concentrations improves the applicability of our model but also introduces uncertainty into our results. A comparison between modelled and measured NO 2 and O 3 background concentrations yields deviations between +13% and −63% for NO 2 and between −10% and +51% for O 3 (Table 5), suggesting that our modelling can indeed provide a robust first order approximation of the NO 2 air pollution for various NO X emission scenarios. These relatively limited deviations also suggest that the correlation function applied between the modelled and measured daily average NO 2 and O 3 concentrations is generally applicable. Depending on the cause of the differences between modelled and measured component concentrations, this generic correction may be too high or low. If the differences stem from factors that are unrelated to traffic emissions, no correction should be applied. If the differences stem from errors in the chemical transport model a proportional correction should be applied. The correction based on a correlation as implemented here constitutes a practical compromise between the two.
In cities like Paris and London, the dimensions of typical street canyons may be larger than the ones assumed here, i.e. streets may comprise 4 lanes in each direction. The NO 2 concentrations in such streets are likely to be higher than those modelled here, as the additional NO X vehicle emissions tend to over-compensate atmospheric dilution and mixing effects of a wider street canyon. Our modelling may thus systematically underestimate NO 2 pollution in the largest streets. The effects of the 10 NO X emission scenarios are, however, similar for larger and smaller streets; consequently the principal conclusions of our modelling should remain valid for large artery streets as well.

Effect of NO X emission levels and NO 2 fractions on urban NO 2 concentrations
In line with the findings of , our analysis demonstrates that overall NO 2 concentrations in the 8 European cities addressed here depend to a larger extent on the NO X emission levels of diesel cars than on the NO 2 fractions (Table 3; compare Scenario 5 with Scenarios 7 and 10). The impact of NO 2 fractions on the ambient NO 2 concentrations is, however, relevant in bigger cities with a large contribution from traffic to total NO 2 pollution as it is the case, e.g., in Paris. As these observations suggest, reductions in the NO X emissions from diesel cars can substantially reduce the annual average NO 2 concentrations. This can be achieved with cleaner diesel cars, a shift to more gasoline cars in city centres or a reduction of motorised traffic. The European Commission has recently introduced the Real-Driving Emissions (RDE) test procedure to measure and control the NO X emissions of diesel cars on the road. Whether this measure will effectively limit the NO X emissions of diesel cars remains to be seen.
Moreover, a reduction in the NO X emissions of diesel cars may not be equally effective in addressing air quality problems in the various European cities. For cities like Athens in which diesel cars contribute little to air quality problems, complementary measures that also address the regional NO 2 background concentrations may need to be considered.
Our modelling results support the findings of Kiesewetter et al. (2014) who modelled the NO 2 pollution at 1950 air quality monitoring stations across Europe. Nevertheless, their method and the one we applied differ in key features. While we model the NO 2 concentrations in a street canyon by accounting for the NO X emissions in the canyon and the NO 2 background concentrations, Kiesewetter et al. (2014) derive the parameters (residence time in the canyon, and emissions) of the box model which computes the traffic increment from NO 2 measurements for the base year. In this way all European measurement stations can be considered, while our approach can only be applied to selected cities due to limited data availability and computational power.
Another difference is related to the impact of variations in NO 2 fraction on the background NO 2 concentrations. Kiesewetter et al. (2014) assume this ratio to be constant whereas we, on the other hand, account for variations of the ratio in the background with the chemical transport model (LOTOS-EUROS; Schaap et al., 2008). Finally, the method applied by Kiesewetter et al. (2014) can only model the annual average NO 2 concentrations and captures thus only the exceedance of annual mean NO 2 limits can be studied whereas we can address here also hourly NO 2 concentrations. This limitation of the approach chosen by Kiesewetter et al. (2014) is not much of a problem since almost all stations that exceed the hourly NO 2 limit on a regular basis, also exceed the annual NO 2 limit. However, if the NO 2 ratios of the NO X emissions continue to increase, exceedances of the hourly NO 2 limit could become more critical from the perspective of air quality.
As a consequence of the dieselgate scandal the French (UTAC, 2016), German (BMVI, 2016) and UK (Department for Transport, 2016) type approval authorities published a report on emissions of Euro 5 and 6 passenger cars. For a full discussion on these tests see Degraeuwe and Weiss (2017). They tested 78 Euro 6 diesel passenger cars, selecting the most common vehicles on the market. All vehicles comply with the NEDC on the test bench but when the same vehicle is driven on the road average emissions are 377 mg NO X /km (4.7 times the limit). Only 16% of the Euro 6 cars tested complied under on-road testing with the NEDC, by emitting less than 80 mg NO X /km, which demonstrates that it is technically possible to comply with the limit on the road. During the onroad test the ambient temperature was 10°C (σ = 4°C). Manufactures argue that this is because cars use legal defeat devices that reduce NO X abatement below 20°C. They claim that this is necessary to protect the engine. However, the legislation also specifies that cars should not reduce after-treatment under normal conditions of use. The legal aspects are outside the scope of this paper. But the outcome of these reports has implications for urban air quality. The normal ambient temperature at which cars operate in Europe is 12°C (σ = 4°C; Malfettani et al., 2016). This means that the results are representative for normal conditions of use in Europe. Also the Committee of Inquiry into Emission Measurements in the Automotive Sector of the European Parliament investigated the discrepancy between the official and on-road NO X emission of diesel cars and published its final conclusions in April 2017 (Gieseke and Gerbrandy, 2017). They conclude that not the supposedly unrealistic NEDC but rather the use of defeat devices explains the biggest part of the difference between emissions on the road and on the NEDC (conclusions 1 and 5, Gieseke and Gerbrandy, 2017). Concerning the conformity factors, allowing diesel cars to exceed the limits by a factor 2.1 after 2019 and a factor 1 plus measurement uncertainty (currently set at 0.5) after 2021, the committee states that it allows the continuing use of less efficient technology. The technology to achieve the Euro 6 standard on the road is available but is not used for economic reasons. Hence, cities that want to tackle the exceedances of the NO 2 limits cannot count on clean diesels arriving soon. They will have to implement alternative measures such as a ban on diesel cars, a reduction of traffic, a shift to gasoline and electric vehicles or a modal shift to cycling, walking and public transport.

Conclusions
We generalise in this article the NO 2 pollution model developed by  and apply it to 8 cities in Europe. Benchmarking our results against measurements of actual NO 2 concentrations suggests that our modelling approach yields a reliable first-order approximation of the effect of various NO X emission sources on urban NO 2 concentrations in European cities. We draw the following conclusions from our research: • The absolute level of traffic NO X emissions is more relevant for the NO 2 concentration in urban street canyons than the NO 2 fraction in the total NO X emissions for most cities. A reduction in the NO X emissions per km of diesel cars can decrease the ambient NO 2 concentrations in European cities.
• In cities with a low share of diesel cars in the vehicle fleet (e.g., Athens) or a high contribution to urban NO 2 pollution from background NO 2 concentrations (e.g., Krakow), additional measures addressing heavy-duty vehicles, and NO 2 sources in the manufacturing, energy, and mining industry are necessary to reduce urban air pollution.
• A sensitivity study shows that the NO 2 background concentration, the wind speed and the diesel share of the fleet have the biggest influence on the NO 2 concentration in a street canyon.
• The NO 2 fraction of NO X emissions only has an effect close to the emission sources and has an impact on the exceedance of the hourly limit value of 200 μg/m 3 . It could become a concern if cars continue to emit a lot of NO x in combination with a higher NO 2 fraction.
• Compliance of diesel cars with the Euro 6 NO x emission limits under real-world driving conditions yields substantial benefits for air quality and can effectively reduce the current exceedances of the NO 2 air quality standard. However, if Euro 6 diesel cars continue to emit on the road substantially more NO X than permitted by the Euro 6 emissions limit, policy makers should consider alternative measures including forcing a shift to gasoline cars to achieve compliance with air quality standards.
• As a consequence of the dieselgate scandal a lot of data on real driving emissions of Euro 6 cars has become available. The on-road performance of the bulk of these cars is not good enough to solve the NO 2 problem in major cities. The allowed conformity factors will delay the widespread adoption of available emission control technologies. Cities with a severe NO 2 problem, may be obliged to consider alternative measures including reducing the share of diesel cars in the urban fleet.

Disclaimer and Acknowledgments
The views expressed here are those of the authors and may not be regarded as the official position of the European Commission. This research was commissioned by the European Commission's Directorate General GROW. The authors would like to thank Martijn Schaap, Carlijn Hendriks and Richard Kranenburg (TNO) for the calculations with LOTOS-EUROS, Stefan Hausberger (TU Graz) and Norbert Ligterink (TNO) for their advice on Euro 6 Diesel emission factors, Hans Hooyberghs for the comparison with the IIASA methodology, and the anonymous reviewers for their comments on an earlier draft of this article.