Report to European Commission Directorate-General Environment

Transcription

1 The Impact of Changes in Vehicle Fleet Composition and Exhaust Treatment Technology on the Attainment of the Ambient Air Quality Limit Value for Nitrogen Dioxide in 21 Report to European Commission Directorate-General Environment Susannah Grice John Stedman Andrew Kent Melanie Hobson John Norris John Abbott Sally Cooke ED48527 AEAT/ENV/R/244 Issue 2 May 27

2 Title Customer Customer reference Confidentiality, copyright and reproduction File reference The Impact of Changes in Vehicle Fleet Composition and Exhaust Treatment Technology on the Attainment of the Ambient Air Quality Limit Value for Nitrogen Dioxide in 21 European Commission Directorate General Environment Service contract 751/26/438919/MAR/C3 This report is the Copyright of AEA Technology and has been prepared by AEA Technology plc under contract to the European Commission. The contents of this report may not be reproduced in whole or in part, nor passed to any organisation or person without the specific prior written permission of the Commercial Manager, AEA Technology plc. AEA Technology plc accepts no liability whatsoever to any third party for any loss or damage arising from any interpretation or use of the information contained in this report, or reliance on any views expressed therein. X:\report\euno2report_maintextv15.doc; X:\report\euno2report_appendicesv5.doc Reference number AEAT/ENV/R/244 Issue 2 AEA Energy & Environment The Gemini Building Fermi Avenue Harwell Didcot OX11 QR t: f: [email protected] AEA Energy & Environment is a business name of AEA Technology plc AEA Energy & Environment is certificated to ISO91 and ISO141 Author Name Susannah Grice John Stedman Andrew Kent Melanie Hobson John Norris John Abbott Sally Cooke Approved by Name Tony Bush Signature Date 2/7/27 AEA Energy & Environment ii

3 Executive summary Objective The object of this study is to examine the impact of changes in primary NO 2 emissions from road vehicles on current and future ambient NO 2 concentrations in EU Member States within the context of the achievability of the 21 ambient air quality limit values for NO 2. Background Two air quality limit values have been established for NO 2 (nitrogen dioxide) in ambient air. The first is an annual value of 4 μg m -3 and the second is an hourly value of 2 μg m -3 with 18 permitted exceedences each year. These limits will enter in force on 1 January 21 and will apply everywhere including roadside locations. Thermal combustion processes emit a mixture of nitrogen oxides (NO X ) in the form of NO (nitric oxide) and NO 2. European legislative standards have set limits for the tail-pipe emissions of NO X from road vehicles (Euro standards). In most ambient locations the majority of NO 2 present in air is formed by oxidation of emitted NO which generally has been viewed as the dominant component of emitted NO X. However, the proportion of NO X emitted from road vehicles directly as NO 2 (f-no 2, often expressed as a percentage) can have a significant impact on ambient NO 2 concentrations, particularly at the roadside, close to the point of emission. Diesel vehicles generally emit more NO 2 than petrol vehicles. Recently, there has been pressure to fit diesel vehicles with exhaust after treatment technology, such as particulate traps and oxidation catalysts in order to meet the emission limits for particulate matter and other air pollutants. In conjunction with an increase in the proportion of diesel-engine vehicles in national fleets this has resulted in an increase in the fraction of NO X emitted as primary NO 2. Thus, there is increasing concern that Member States may experience difficulty complying with the annual mean limit value for NO 2 of 4 μg m -3. Recent trends in f-no 2 Measurements of the f-no 2 from different road vehicles have not been undertaken within this study but relevant information has been summarised. From the data reviewed, it is clear that the previously accepted assumption of a 5% f-no 2 rate is a systematic underestimate for diesel vehicles. Furthermore, the data strongly indicates that no single value of f-no 2 is appropriate for all vehicle types. Rather, it is dependent on: Vehicle type (passenger car, heavy goods vehicle, bus etc) Emission standard Any exhaust after treatment fitted The average speed of the drive cycle The data for petrol fuelled vehicles shows f-no 2 has remained around 3 4% for all technologies and emissions standards. For older diesel vehicles f-no 2 is around 11%. However, this value changes with changes in vehicle technology. In particular oxidation catalysts, either used alone or on the engine side of particulate traps, lead to f-no 2 increasing to around 3%. NO X reduction technologies, whilst leading to a reduction in NO X, typically of between 3 and 5%, also lead to an increase in f- NO 2, with ratios up to 6% being observed for some new passenger car technologies. However, for heavy-duty vehicles the introduction of selective catalytic reduction (which is the technology favoured by the majority of engine manufacturers to meet Euro IV emission standards) appears to reduce NO X and NO 2 emissions with f-no 2 having been measured around 5% to 1 % albeit for a small number of vehicles. An understanding of the recent trends in vehicular f-no 2 emissions in Europe is best obtained from a combined examination of ambient monitoring data from individual monitoring sites and the compilation of emission inventories. This is the approach we have adopted, which enables a comparison to be made of these independent estimates of primary NO 2 emissions. This approach also provides us with the tools to make predictions of ambient NO 2 concentrations in future years. AEA Energy & Environment iii

4 An examination of recent trends in the ratio of ambient NO 2 and NO X concentrations at roadside locations does not provide us with a clear picture of the change in f-no 2 directly. This is because NO X emissions and thus ambient NO X concentrations have generally declined as a result of the Euro standards and this leads to an increase in the ratio of ambient NO 2 to NO X concentrations, due to a shift in the equilibrium between ozone, NO and NO 2 concentrations. Thus, a model is required to distinguish between these influences and those arising from changes in f-no 2. In this study we have used the Netcen Primary NO 2 model for this purpose. Ten case study locations (listed in Table E1) were chosen for detailed analysis in this study. These locations were chosen to provide a reasonable geographical coverage of the EU with an emphasis on locations with the highest roadside NO 2 concentrations. Estimates of f-no 2 for recent years have been calculated for selected monitoring sites in each location using the Netcen Primary NO 2 model. Estimates of average national and national urban values of f-no 2 have also been calculated using data on fleet characteristics and NO X emission factors within the TREMOVE model and a summary of emission factors for primary NO 2 from different vehicle classes, Euro standards and technologies compiled for this study. Table E1 Details of case study countries and cities Country City (if applicable) Austria Salzberg/Hallein Czech Republic Prague Finland - France Paris Germany Baden Wűrttemberg Greece Italy Netherlands - Spain UK Athens Milan Barcelona London The results of the analysis of recent trends in f-no 2 show that f-no 2 has increased over the past five to ten years in the majority of countries considered here. The emissions results and the results of the Netcen Primary NO 2 modelling indicate that the rate of increase in f-no 2 has generally increased since 2 compared with the rate of increase in f-no 2 between 1995 and 2. A comparison of the local scale Netcen Primary NO 2 model results and the national scale emissions results show that local factors (e.g. characteristics of individual roads) can have a significant impact on the f-no 2. Evidence for this can be seen in the scatter of f-no 2 modelling results for different roads within a single case study area. Additionally, regional factors can impact on f-no 2. For example both in Baden Wűrttemberg and London, a greater number of buses have been fitted with particulate traps than the national average and these local differences from the national average in terms of vehicle fleets may have caused f-no 2 to be above the average as calculated using national scale emissions analysis. The comparison of f-no 2 estimates are summarised in Table E2. This table presents both the emission inventory based estimates of f-no 2 for urban areas for the baseline scenario and the range of f-no 2 estimates derived from monitoring data (see Table E3. for details of the emissions scenarios used in this report). The reason for showing a range rather than an average value is we have only considered 4-6 roadside monitoring sites within each case study area. Therefore, while these sites do represent a range of local conditions, they are not necessarily representative of a wide area. AEA Energy & Environment iv

5 Table E2. Summary of f-no 2 (percent) for each member state derived from both ambient measurements and emission inventory calculations (baseline scenario) Country Method Austria Czech Republic Finland France Germany Greece Italy Netherland s Spain UK Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Baseline emissions projections and air quality Estimates of f-no 2 have also been calculated using an enhanced version of the TREMOVE emission inventory for 21, 215 and 22. The Netcen Primary NO 2 model has then been used to predict ambient NO 2 concentrations at the monitoring sites in the case study areas. The baseline projections of road transport emissions take into account the recently agreed Euro 5 and 6 regulations for light duty vehicles and the Euro V stage of emissions control for heavy duty vehicles. They do not include any assumptions about a possible Euro VI for HDVs. NO X emissions are generally predicted to decrease steeply from 25 to 22. This trend is apparent in both urban and national scale emissions. f-no 2 is generally predicted to increase steeply from Thus urban NO 2 emissions are predicted to increase from 2 to 21 in contrast to the decline in NO X emissions. Urban NO 2 emissions are then predicted to flatten off to 215 and then decline to roughly equivalent 25 values in 22 for the baseline. By 22 the decrease in NO X emissions is sufficient to offset the increase in f-no 2. Baseline annual mean NO 2 concentrations are predicted to decline between 25 and 22 at most of the case study locations. At many of the sites considered the annual mean NO 2 limit value is expected to be exceeded in 21. At a significant number of these sites the annual mean concentration is expected to remain above the limit value until 22 and beyond. Little or no decline is predicted between 25 and 21 at the case study sites in the Czech Republic and Spain, presumably due to a combination of relatively modest decreases in NO X emissions from traffic and an increase in f-no 2 at these locations. An increase in annual mean NO 2 concentration between 25 and 21 is predicted at the case study sites in France, due to the high roadside NO X concentrations at these sites and high predicted f-no 2 in 21. In all other locations the predicted increase in f-no 2 is not large enough to offset the decrease in NO X emissions. AEA Energy & Environment v

6 A range of sensitivity analyses have also been carried out for the baseline predictions of annual mean NO 2 concentrations. These include: Projections for 21, 215 and 22 calculated without taking changes in f-no 2 into account, tend to under predict concentrations in the future. This has the greatest impact at roadside locations with the highest NO X concentrations. Projections for 21, 215 and 22 calculated with different assumptions made about changes in regional oxidant into the future. The magnitude of the impact on annual mean NO 2 concentrations is generally smaller than the impact of changing f-no 2, particularly at sites with the highest measured concentrations. Scenarios Ambient NO 2 concentrations have been predicted in 21, 215 and 22 for four scenarios in addition to the baseline. These scenarios have been used to examine the impact of different options for future emission limits and exhaust after treatment technologies on estimates of ambient NO 2 concentrations. The scenario assumptions are listed in Table E3. Table E3. Summary of NO X emissions and f-no 2 scenarios Scenario Description baseline NO X projections include the impact of Euro 5 and Euro 6 Light Duty Vehicles (LDV). Baseline f- NO 2 scenario 1 NO X projections include the impact of Euro 5 and Euro 6 (LDV). More pessimistic f-no 2 for Heavy Duty Vehicles (HDV) than the baseline scenario 2 NO X projections include the impact of Euro 5, Euro 6 (LDV) and an estimate for Euro VI (HDV) based on US standards. More optimistic f-no 2 for HDV than the baseline scenario 3 NO X projections include the impact of Euro 5 and Euro 6 (LDV). More optimistic f-no 2 for LDV at Euro 6 than the baseline (scenario 3 Euro 6 LDV f-no 2 of 1% rather than 55% for the baseline) scenario 4 NO X projections include the impact of Euro 5, Euro 6 (LDV) and an estimate for Euro VI (HDV). Same f-no 2 and NOx control assumption for HDV as scenario 2 and more optimistic f-no 2 for LDV than scenario 3 Figure E1 summarises the emission projections for the different scenarios in terms of total NO X and NO 2 emissions and average f-no 2 across the ten member states considered. NO X emissions decline steeply to 22 and are lowest for scenarios 2 and 4 in 22. f-no 2 is predicted to increase steeply from 25 to 215 and then starting to decline for scenarios 3 and 4. As a result, NO 2 emissions are predicted to increase from 2 to 21 in contrast to the decline in NO X emissions. For the baseline and scenarios 1 and 2, NO 2 emissions are then predicted to flatten off to 215 and then decline to roughly equivalent to 25 values in 22. By 22 the decease in NO X emissions is sufficient to offset the increase in f-no 2. NO 2 emissions are projected to decrease more steeply to 22 to values below 2 emissions if there were lower emissions of primary NO 2 from light duty vehicles (scenario 3) combined with new standards for heavy duty vehicles (scenario 4). AEA Energy & Environment vi

7 Figure E1 Graphs summarising urban road traffic NO x and NO 2 emissions summed across the ten member states considered along with average urban f-no 2 for the different emission projection scenarios a) NO X emissions (ktonnes per year) b) NO 2 emissions (ktonnes per year) Baseline Scenario 1 Scenario 2 Scenario 3 Scenario Baseline Scenario 1 Scenario 2 Scenario 3 Scenario ktonnes per year ktonnes per year year year c) f-no 2 (percent) 4% 35% 3% Baseline Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (percent) 25% 2% 15% 1% 5% % year The predictions of annual mean NO 2 concentration are found to be most sensitive to the scenario assumptions at the sites with the highest NO X concentrations. As expected the predicted NO 2 concentrations for scenario 1 are higher than the baseline since f-no 2 for heavy duty vehicles is assumed to be higher than the baseline. The predictions for scenarios 2 and 3 are similar being lower generally than the baseline but with some site to site variation. Thus the impact of the reductions in NO X and f-no 2 for heavy duty vehicles in scenario 2 result in roughly equivalent ambient NO 2 concentrations to the impact of the reduction in f-no 2 for light duty vehicles assumed for scenario 3. The lowest ambient NO 2 predictions were obtained for scenario 4, which incorporates the reduction in NO X emissions and f-no 2 for heavy duty vehicles from scenario 2 with an additional reduction in f-no 2 for light duty vehicles relative to that assumed in scenario 3 (6% f-no 2 for Euro 6 cars and Euro 5 and 6 vans, compared with 1% assumed in scenario 3 and 55% in the baseline). AEA Energy & Environment vii

8 Typical Impact of Changes in the Annual Mean NO 2 as a Result of Changes in f-no 2 The site-specific results from the Netcen Primary NO 2 model have been analysed further with a view to drawing some general conclusions regarding the likely impact of future changes in primary NO 2 emissions on the achievability of the annual mean limit value of 4 μg m -3 at the roadside in Member States. We have approached this by calculating the difference (Delta NO 2 1) between the site-specific predictions for the baseline and the sensitivity test in which f-no 2 has been fixed at 25 values. Regression analysis has been used to derive the relationship between Delta NO 2 1 and the baseline annual mean NO 2 projections. The results of this analysis suggest that the changes in f-no 2 implied by the baseline will result in approximately an additional 2 μg m -3 annual mean NO 2 at 4 μg m -3 and μg m -3 at 6 μg m -3 at roadside locations in 21. The impact is expected to be greater in 215 and 22 with an additional 4 μg m -3 at 4 μg m -3 and μg m -3 at 6 μg m -3. In future work investigating f-no 2 the use of local scale emissions data in generating f-no 2 projections is recommended. There are local inventories available, for example the London Atmospheric Emissions Inventory (GLA, 26), which contain data on local vehicle fleets, locally specific vehicle trip characteristics and road networks. This could be complemented with more comprehensive measurements of f-no 2 from specific vehicle types and emerging technologies. Concluding remarks The impact of primary NO 2 emissions on ambient air quality is expected to be most pronounced at roadside locations with the highest NO X concentrations. A combination of emission inventory calculations and projections of ambient air quality suggest that the impact on ambient air quality will be greatest between 25 and 215. Changes in vehicle exhaust after treatment technology, particularly selective catalytic reduction, are expected to result in a decrease in the emissions of primary NO 2 and an improvement in roadside air quality by 22. The magnitude of primary NO 2 emissions, NO 2 concentrations and the extent of exceedence of the annual mean limit value in 22 will be dependent on the exhaust limit values set, the technology adopted and the performance of the technology, as illustrated by our scenario calculation results. At many of the sites considered the annual mean NO 2 limit value is predicted to be exceeded in 21. At a significant number of these sites, the annual mean concentration is expected to remain above the limit value until 22 or beyond even with new Euro VI standards to address the emissions from heavy duty vehicles. AEA Energy & Environment viii

9 Table of contents 1 Introduction Policy Context Report Structure Project Requirements The Netcen Primary NO 2 Model 3 2 Ambient Monitoring Data Case Study Locations Monitoring Data Summary Statistics Summary 2 3 Recent Trends in f-no 2 using the Netcen Primary NO 2 model Introduction Ozone Module Verification Best Estimate f-no 2 Results f-no 2 Sensitivity Analysis 25 4 Emissions Analysis Introduction NO X Type Approval Limits Estimating NO X Emissions Estimating NO 2 Emissions Scenarios Summary 55 5 Comparison of f-no 2 from the Netcen Primary NO 2 model and Emissions Analysis for Recent s Introduction Results Comparison Conclusions from Analysis of Recent Trends in f-no Future Ambient NO 2 Concentrations: Baseline projections Introduction Modelling Assumptions Projection Module Verification Baseline Modelling Results Conclusions 79 7 Future Ambient NO 2 Concentrations: Baseline Sensitivity Analysis Introduction Model Sensitivity to f-no Model Sensitivity to Future Regional Oxidant Levels 84 AEA Energy & Environment ix

10 7.4 The Impact of Under Predicting Base Model Results on Modelled Future NO 2 Concentrations Conclusions of Sensitivity analysis 91 8 Future Ambient NO 2 Concentrations: Scenario Projections Introduction Modelling Assumptions and Emissions Projections Scenario Modelling Results 93 9 Wider Applicability of Model Results Across the EU Introduction An Analysis of the Typical Impact of Changes in the Annual Mean NO 2 as a Result of Changes in f-no Case Study: An Assessment of the Extent of Exceedences of the Annual Mean NO 2 Limit Value across the UK Relevance of Typical Impacts to Estimate Extent of Exceedences in the UK and Europe Concluding remarks Acknowledgements References 122 Appendices Appendix 1 Appendix 2 Appendix 3 Ambient Monitoring Data Ozone Module Verification Analysis of Primary Oxidant from Road Transport and Secondary Background Oxidant AEA Energy & Environment x

11 AEA Energy & Environment xi

12 1 Introduction 1.1 Policy Context European legislative standards controlling the pollutants released in vehicle exhaust gases, such as Euro IV and the forthcoming Euro V for heavy duty vehicles (HDV) are intended to reduce the total emissions of NO X from vehicle exhausts without differentiating between NO and NO 2 fractions. In contrast, the Framework Directive (Directive 96/62/EC) and First Daughter Directive (Directive 1999/3/EC) have been designed to control the concentrations specifically of NO 2 in ambient air to which the public is exposed. Details of the limit values for NO 2 specified in the First Daughter Directive are given by Table 1.1. These two different approaches to controlling oxides of nitrogen in air have resulted in a legislation gap whereby vehicle manufacturers have reduced NO X emissions in compliance with the Euro standards and other directives but this has not yielded a reciprocal reduction in NO 2 levels in ambient air sufficient to meet limit values. Table 1.1. Limit values for NO 2 specified in the First Daughter Directive. Averaging period Limit Value Comes into force ly 4 μg m -3 1/1/21 Hourly 2 μg m -3 (18 exceedences allowed per year) 1/1/21 To compound this, the proportion of direct NO 2 emissions from vehicle exhausts may be rising as a result of changes in the composition of national vehicle fleets across Europe and the introduction of new exhaust technologies that have been introduced to meet the emission limits for various pollutants. For petrol-fuelled vehicles the proportion of NO X emitted directly as NO 2 is less than 5%, whereas this proportion in diesel vehicles not fitted with new exhaust treatment technology is higher at around 1-12%. The continuing increase in the proportion of diesel-engine vehicles in national fleets will therefore have a significant impact on the concentration of ambient NO 2 levels, particularly at roadside environments. Furthermore, the pressure to fit diesel vehicles with after exhaust treatment technology such as particulate traps and oxidation catalysts is likely to further increase the proportion of NO X emitted as NO 2. Some catalyst-based particulate filters achieve the catalytic action by oxidising a portion of the NO in the exhaust to NO 2 in order to promote the oxidation of soot collected in the filter and so potentially emit a higher proportion of NO X as primary NO 2. There are also other pressures besides advances in abatement technology and future exhaust standards to consider. The automotive industry s pursuit of increased fuel economy using lean burn technology such as gasoline direct injection (GDI) produces a higher air to fuel ratio and more oxygen in the burnt mixture resulting in increased NO X from the tailpipe. As a result of these changes there is an increasing interest in the primary NO 2 fraction (f-no 2 ), defined as the fraction of NO X emitted as NO 2, from different types of vehicles (often presented as a percentage). More attention is now being placed on assessing the impact of these emission changes on ambient levels of NO 2, particularly at roadside locations where human exposure can be significant. Both future changes in vehicle fleet emissions as current and future exhaust standards take force and changes in abatement technologies will have a significant impact on future ambient NO 2 concentrations. It will therefore be necessary to use information on these changes to model future ambient NO 2 concentrations at roadside locations. Although current concerns with the NO 2 limit values mainly relate to compliance with the 21 annual limit value, changes in the proportion of directly emitted NO 2 could influence the relative stringency of the objectives. The 21 annual limit value is generally considered to be the more stringent of the two NO 2 limit values so the project is focussed on assessment of the annual limit value. However, this might not necessarily continue to be the case as primary NO 2 increases so additional consideration is also given to the hourly limit value. An understanding of the recent trends in f-no 2 in vehicle emissions in Europe is best obtained from a combined examination of ambient monitoring data and the compilation of emission inventories. This is the approach we have adopted in this study as it enables a comparison to be made between these AEA Energy & Environment 1

13 independent estimates of primary NO 2 emissions. This combined approach also provides us with the tools to make predictions of ambient NO 2 concentrations in future years. An examination of recent trends in the ratio of ambient NO 2 and NO X concentrations at roadside locations does not necessarily provide us with a clear picture of the change in f-no 2 directly. This is because NO X emissions and thus ambient NO X concentrations have generally declined as a result of the Euro standards and this leads to an increase in the ratio of ambient NO 2 to NO X concentrations due to a shift in the equilibrium between ozone, NO and NO 2 concentrations. As a result a model is required to distinguish these influences from changes in f-no 2. In this study we have used the Netcen Primary NO 2 model as described below. 1.2 Report Structure The objective of this work is to assess how changes f-no 2, caused by changes in vehicle fleet composition and exhaust treatment technology, will impact on the compliance with the NO 2 limit values set out in the First Daughter Directive which comes into force in 21. To meet this objective, the report will be divided into two main sections. The first section will cover recent trends in f-no 2 including: Analysis of ambient monitoring data (chapter 2) Modelling f-no 2 using the Netcen Primary NO 2 model (chapter 3) Emissions based calculations of f-no 2 (both recent trends and future projections), largely based on Tremove (chapter 4) Comparison f-no 2 estimates for recent years calculated using the Netcen Primary NO 2 model and f-no 2 from emissions analysis (chapter 5) The second section will focus on the likely impact of changes in f-no 2 on future ambient NO 2 concentrations. This will include the following: Calculating baseline projections of ambient NO 2 concentrations with the Netcen Primary NO 2 model using f-no 2 projections from chapter 4 (chapter 6) Sensitivity analysis on some of the key assumptions made in generating the baseline NO 2 projections (chapter 7) Emissions scenario modelling (chapter 8) Extrapolation of the model results across the EU (chapter 9) An assessment of the achievability of the limit values set out in the First Daughter Directive as a result of changes in f-no 2. Chapter 1 will summarise our conclusions regarding this. Table 1.2 summarises the model runs carried out as part of the assessment of future ambient NO 2 levels. Predictions of future ambient NO 2 concentrations are highly dependent on future Euro standards and the technology adopted to meet these standards. We have therefore examined a range of possible future scenarios including our best estimate of the impact of current legislative proposals and various different possible future emissions standards. AEA Energy & Environment 2

14 Table 1.2. Summary of NO X emissions and f-no 2 scenarios included in the assessment of future ambient NO 2 concentrations Scenario Description baseline NO X projections include the impact of Euro 5 and Euro 6 on Light Duty Vehicles (LDV). Baseline f- NO 2 scenario 1 NO X projections include the impact of Euro 5 and Euro 6 (LDV). More pessimistic f-no 2 for HDV than the baseline scenario 2 NO X projections include the impact of Euro 5, Euro 6 (LDV) and an estimate for Euro VI (HDV) based on US standards. More optimistic f-no 2 for HDV than the baseline scenario 3 NO X projections include the impact of Euro 5 and Euro 6 (LDV). More optimistic f-no 2 for LDV at Euro 6 than the baseline scenario 4 NO X projections include the impact of Euro 5, Euro 6 (LDV) and an estimate for Euro VI (HDV). Same f-no 2 and NOx control assumption for HDV as scenario 2 and more optimistic f-no 2 for LDV than scenario 3 For the analysis outlined above, a case study approach has been adopted whereby ten locations from across the EU have been considered in detail. These locations have been selected to represent a range of different geographical conditions (including different ozone climates) and to reflect a range of the different vehicle fleet compositions found in the EU. Some consideration is then given to assessing the extent to which these results can be generalised across the EU as a whole. Further details of countries and cities considered are included in chapter Project Requirements There are four distinct tasks in this project. Details of these tasks and how they correspond to chapters in this report are given in Table 1.3. Task 1 is primarily concerned with understanding the current situation with regards to the proportion of NO X that is directly emitted as NO 2 (f-no 2 ). This can either be expressed as a proportion or as a percentage. As part of understanding the current situation we have looked at recent trends in f-no 2 time series data over a 5-1 year period. Tasks 2, 3 and 4 are concerned with understanding how these and future changes in f-no 2 are likely to impact on future ambient NO 2 concentrations. This involves using model projections to assess the likely compliance date for meeting the 21 limit values for NO 2. It also involves looking at the sensitivity of the model projections to changes in f-no 2 and the likely impact of new vehicle emissions standards on projected NO 2 concentrations The Netcen Primary NO 2 Model Much of the analysis presented in this report relies upon the Netcen Primary NO 2 model. A brief description of this model, including its three component modules, is given in Box 1.1. A fuller description of the model and its application to monitoring sites within the UK can be found in Abbott (26). AEA Energy & Environment 3

15 Table 1.3. How the four project tasks are covered by this report Task 1A. Compile and review any publicly available information relating to measured emissions of nitrogen dioxide and total NO X from the various vehicle categories and age across the EU. 1B. Where possible, the contractor should use the information obtained to assess the evolution of absolute primary emissions of nitrogen dioxide and their proportion relative to total emissions of nitrogen oxides from a range of vehicle types in the EU Corresponding report chapter Chapter 4 includes details of a number of studies of NO 2 emissions that have been used to help inform f-no 2 figures to be used in this project. Chapter 2 and 3 present ambient monitoring data and f-no 2 estimates generated using the Netcen Primary NO 2 model for sites from a number of representative locations across the EU. This data covers a 5-1 year period in the run up to 24 and includes numbers for 25 where this data has been readily available. This data is local scale with concentrations and derived f-no 2 values representative of the specific roads for which modelling has been carried out. Chapter 4 presents emissions based estimates of f-no 2 for 1 countries. This data is national scale - we have calculated average national f-no 2 values and average f-no 2 for urban areas in each of these countries. f-no 2 estimates have been calculated for 1995, 2, 25, 21, 215 and 22. Chapter 5 compares these two alternative methods of calculating f-no 2 and draws some general conclusions about how f-no 2 trends have evolved. 2. Using emissions inventories (e.g. European, National or city), air quality monitoring information, air quality modelling or empirical assessments, assess the degree to which the annual average ambient air quality standard for nitrogen dioxide is likely to be exceeded in 21 and 215 within the EU. Chapter 6 presents details of our baseline projections of ambient NO 2 concentrations for 21, 215 and 22 generated using the Netcen Primary NO 2 model. This uses the base year ambient monitoring data from chapter 2 and f-no 2 projections from emissions data calculated in chapter 4. A comparison of the modelled concentrations and the limit values is also presented. The modelling provides local scale concentration estimates. Chapter 9 presents details of how we have attempted to generalise our local scale results across the wider EU area. 3. Assess the sensitivities of the conclusions reached under (2) above, to changes in the proportions of directly emitted nitrogen dioxide in vehicle exhausts. 4. Assess the extent and timeframe to which the predicted air quality exceedences would be reduced by new vehicle emissions standards for nitrogen oxide (NO X ) emissions introduced after 21 (for all vehicle types) as well as the implementation of measures on stationary combustion sources. Chapter 7 presents sensitivity analysis of our model projections for 21, 215 and 22 including the impact of changing f-no 2. Chapter 8 presents modelled ambient NO 2 concentrations for 21, 215 and 22 for four emissions scenarios. Analysis of the impact of these scenarios on the likely timeframe for compliance with the limit values is also presented. AEA Energy & Environment 4

16 Box 1.1 The Netcen Primary NO 2 Model The Netcen Primary NO 2 Model is a one-dimensional model of the interaction between the primary NO 2 ratio and NO X, NO 2 and O 3 concentrations at roadside locations. It has been developed and used within the UK Defra air quality research programme. It is a local scale model. This makes it appropriate for analysis of the primary NO 2 ratio (f-no 2 ) and NO 2 concentrations for compliance with limit values at roadside locations because it is computationally efficient at this scale and can pick up the local processes occurring better than a larger scale model could. Several relationships and assumptions under-pin the model. These include: A background site can be chosen to be paired with each roadside monitoring site such that the NO x, NO 2 and O 3 measured at the background site are representative of the background concentrations at the roadside site. Total oxidant at roadside locations [Ox] = [O 3 ] + [NO 2 ]; [Ox] 1 - [Ox] = A ([NO x ] 1 [NO x ] ) + B* where A is the primary NO 2 ratio, Ox is the total oxidant (1 is for roadside, is for background) and B* represents the net effect of other reactions and deposition and excludes the background oxidant concentration. The model can be used in several different forms depending on what input data is available and what information is needed. Three separate modules have been developed for calculating the different parameters as described below: Module 1: The analysis module. This calculates the primary NO 2 ratio for roadside monitoring sites using hourly NO X, NO 2 and O 3 measurements from this site and hourly NO X, NO 2 and O 3 measurements from its paired background site. The annual primary NO 2 component is derived directly from the monitoring data by regressing the hourly roadside increment of oxidant (dependant variable) against the hourly roadside increment of NO X (independent variable). The annual f-no 2 is calculated as the gradient of the regression line. This is useful because primary NO 2 emissions are difficult to measure directly and few measurement studies are available. Using this module of the model will enables us to analyse recent time series trends in the primary NO 2 ratio. Module 2: The ozone module. The ozone concentration at the roadside is calculated using a onedimensional finite difference model of the chemistry and turbulent diffusion in the surface boundary layer. The primary NO 2 ratio is then derived from the monitoring data by regression analysis. There are relatively few roadside monitoring sites across the EU where ozone is measured. This module has therefore been used in the analysis of recent time series data so that we are not limited to choosing roadside sites with ozone monitoring in attempting to select a representative sample of sites across Europe. This module is also used in projecting ozone concentrations into the future for use within the prediction module. Module 3: The prediction module. This module uses the one-dimensional finite difference model to calculate hourly NO 2 and O 3 concentrations at the monitoring site. This is used in modelling future concentrations. In order to run the model for future years, assumptions have to be made about the following parameters. These can be varied as part of the scenario modelling: The primary NO 2 ratio (e.g. does this remain constant from the base year, how much does it increase by?) The met data (e.g. does base year met data apply in future years?) How hourly background concentrations of NO X in the base year will change for future years (e.g. scaling by a factor to represent expressed changes in traffic and non-traffic NO X emissions) How hourly background concentrations of NO 2 in the base year will change for future years How hourly background concentrations of total oxidant in the base year will change for future years How hourly roadside concentrations of NO X in the base year will change for future years The prediction module of the model can be validated by running it for the base year. Comparison between measured and modelled concentrations can then be made. We have validated the prediction module for different sites by running it for the base year. AEA Energy & Environment 5

17 2 Ambient Monitoring Data 2.1 Case Study Locations Ten case study locations from across the EU were selected to be the focus of most of the analysis presented in this report. Table 2.1. presents details of these locations. Table 2.1. Details of case study countries and cities Country City (if applicable) s considered in analysis of recent trends Austria Salzberg/Hallein 2-24 Czech Republic Prague 2-24 Finland France Paris Germany Baden Wűrttemberg Greece Athens 2-24 Italy Milan 2-24 Netherlands Spain Barcelona 2-25 UK London A variety of information sources were used to select the case study locations including: 24 questionnaires submitted by each member state under the First Daughter Directive (CDR, 26) NO X, NO 2 and ozone monitoring data from Airbase (Airbase, 26). Additionally, consideration was given to ensuring that a reasonable coverage of the EU was achieved, including representing a range of ozone climates and both older and more recently joined member states. In Austria, the urban areas of Salzburg and neighbouring Hallein were selected because three sites within them exceeded the annual mean limit value and margin of tolerance for NO 2 in 24, with a maximum exceedence concentration of 58 μg m -3. Prague in the Czech Republic was selected as a case study so that this analysis would include a city from a member state that joined the EU in 24. This is because it is likely that the more recently joined Member States will have significantly different car fleets to the rest of the EU. Prague specifically was selected because there was a site with a reported exceedence of the NO 2 annual mean limit value at 76 μg m -3 in the 24 questionnaire. Finland was chosen as a case study so that the Northern parts of the EU were represented. In Finland, no one city had sufficient monitoring to make it possible to focus on a single city. Therefore, the southern part of the country was selected as the case study area. For Germany, the region of Baden Wűrttemberg was chosen, which contains several urban areas including Stuttgart, Manheim, Karlsruhe and Freiberg. This was because this area contains several sites with high NO 2 concentrations. Stuttgart also has buses fitted with catalyst based particulate filters (Lambrecht, 26), which are likely to have high f-no 2. Additionally, other studies on f-no 2 have already been done in Baden Wűrttemberg using the Carslaw and Beavers (25) modelling approach (Kessler et al, 26). Comparison with this work therefore provides an opportunity to further validate the Netcen Primary NO 2 model. Paris in France was selected as it has a lot of long running monitoring sites. France was considered to be of interest because it is thought that due to the early penetration of diesel vehicle into the French AEA Energy & Environment 6

18 vehicle fleet relative to other Member States, f-no 2 changes may have occurred earlier. For this reason, we have started the analysis in Paris in 1995 to look at a longer time series. A case study area encompassing part of the Netherlands was chosen because the Netherlands have expressed concerns regarding the 21 deadline for meeting the NO 2 limit value. In the Netherlands many of the cities are very close together and inter-linked, so it was decided to broaden the analysis to cover several of these urban areas. Three case study areas within the southern area of the EU were selected to demonstrate whether the model works in the ozone climate associated with this area and therefore to demonstrate the extent to which conclusions drawn in this report apply in the southern EU ozone climate. These areas are: Athens in Greece, Milan in Italy (selected as a case study because the 24 questionnaire showed that many sites in this region of Italy exceeded the annual mean limit value and margin of tolerance in 24) Barcelona in Spain (two sites exceeded the NO 2 limit value and margin of tolerance in 24). London in the UK was selected because f-no 2 has been studied here (see AQEG, 26) and eight sites have been reported in the 24 questionnaire as exceeding the NO 2 annual mean limit value and margin of tolerance. The worst offending site in 24, London Marylebone Road, had an annual mean NO 2 concentration of 11 μg m Monitoring Data Summary Statistics For the analyses of ambient monitoring data and the analyses using the Netcen Primary NO 2 model, between four and six roadside sites have been selected for each case study area. Background site(s) have also been selected to represent background conditions for the case study area. Individual roadside sites have been paired with a nearby background site, which is assumed to represent the non-roadside contribution at the roadside sites. Table 2.2 presents details of the complete list of sites used in this report. None of the sites listed here had closed at the time of writing. Where possible we have selected sites to focus on which have high measured annual mean NO X and NO 2 concentrations. This is because these are the sites where it is most likely that changes in f-no 2 will cause exceedences of the limit values for NO 2. Other considerations in selecting sites included data capture as model results for sites with low data capture will be less reliable than for sites with better data capture. Also sites have generally been selected that have long running data (e.g. >5 years) in order to get the clearest idea possible of trends in f-no 2. However, there are a few sites with very high annual mean NO 2 concentrations that have only opened recently. Where this is the case, we have included these sites in the analysis as well. Because we have selected sites based on these criteria, it would be unrealistic to suggest that they will be representative of the entire case study city/country. However, they should provide a good understanding of changes in f-no 2 at some of the sites most likely to miss the 21 limit values. It is important to look at trends in monitoring data early on in the analysis presented in this report because it is useful to understand as much as possible about the situation is at each monitoring site before modelling f-no 2. Graphs of annual mean concentrations for NO 2, NO X and ozone for the case study sites for 2-25 ( where applicable) are presented in Figure 2.1. NO X concentrations are quoted in μg m -3, as NO 2 throughout this report. Low data capture is highlighted when less than 75%. Complete annual mean statistics including data capture are given in Appendix Austria In the Austrian case study area (Figure 2.1a), four of the five sites considered annual mean concentrations of NO X showed a very slight upward trend between 2 and 23, before concentrations dipped in 24. At the fifth site, AT24A (Hallein Hagerkreuzung), this rise in NO X concentrations was steeper between 2 and 23 and continued in 24. The highest NO X concentrations were found at AT38A (Salzburg Rudolfspaltz) between 2 and 23 while by 24, NO X concentrations at AT24A had increased to the same level as AT38A. NO X AEA Energy & Environment 7

19 concentrations at the roadside site AT168A (Salzburg Mirabellplatz) were not significantly higher than at the background site AT75A (Salzburg Lehen). NO 2 concentrations for sites selected in Austria generally showed a steeper increase between 2 and 23 than for NO X, and similarly to NO X concentrations, there was a dip in 24 at all sites except AT24A where annual mean NO 2 concentrations continued to rise. Two of the four sites exceeded the annual mean limit value for NO 2 for all years considered. The remaining roadside sites showed a similar pattern in NO 2 concentrations to the background site and none of these exceeded the limit value. Only one of the roadside sites, AT168A and the background site AT75A measured ozone between 2 and 24. There seems to be no significant over all trend in ozone, although the peak in 23 caused by the summer heat wave that year is clearly evident in the time series at both sites Czech Republic In Prague (Figure 2.1b), there was little overall change in annual mean NO X concentrations between 2 and 24 at the sites selected for which there are five years of data. However, within this fiveyear period, there was some inter-annual variability. All the roadside sites had noticeably higher NO X concentrations than the background site CZ2A (Pha4-Libus). At CZ66A (Pha2-Legerova), NO X concentrations were much greater than at the other sites in both 23 and 24, although the 23 annual mean concentration only had 42% data capture. For NO 2, at the roadside sites with five years of data, there was an upward trend between 2 and 23, which took NO 2 concentrations significantly above the annual mean limit value by 23. However, in 24, NO 2 concentrations at these sites dropped back down to near the annual mean limit value and lower than the limit value in the case of CZ13A (Pha1-Vrsovice). At CZ66A concentrations significantly above the annual mean limit value were recorded in both 23 (low data capture) and 24. NO 2 concentrations at the background site were significantly lower than the roadside sites selected. Ozone at all sites where it was measured showed a similar overall trend, although the background site had significantly higher annual mean ozone concentrations than at roadside sites. Ozone between 21 and 24 showed an upward trend, with a peak in 23 again due to the summer heat wave Finland The sites selected for analysis in Finland (Figure 2.1c), show little overall trend in terms of annual mean NO X concentrations between 2 and 24: two sites remained constant across this period, one site showed a slight decrease in NO X concentrations and at two sites there was a slight increase. NO 2 annual mean concentrations at all the monitoring sites selected for Finland show an upward trend between 2 and 24, with the exception of the roadside site FI16A (Turun kauppatori) where concentrations have decreased. All sites were below the annual mean limit value for NO 2. Ozone increased marginally between 2 and 24 at the one site at which it was measured France In Paris (Figure 2.1d), annual mean NO X concentrations at all the roadside sites selected decreased between 1995 and 25. Concentrations at the roadside sites were very high compared with other case study areas included in this report with the maximum annual mean NO X concentration in 1995 of 652.3μg m -3 at FR895A (Boulevard périphérique Auteuil). By 25, the annual mean NO X concentration at this site had decreased to 43.8μg m -3, which is still comparatively high. The high NO X concentrations here probably reflect the early uptake of diesel vehicles in France, which would have resulted in higher NO X emissions. Additionally, some of the monitoring sites are located next to very major roads (e.g. FR895A is a kerbside site by Boulevard périphérique). NO 2 concentrations at the selected sites in France again are high compared with other case studies, with annual mean concentrations at FR895A in excess of 1μg m -3 in 23 and 24. Generally, for roadside sites there has been a slight upward trend in NO 2 concentrations until 23 after which, there is a slight decrease. A slight decrease in NO 2 concentrations is evident at the background site FR918A (PARIS 6ème) between 1995 and 25. Ozone at the background site in Paris shows an upward trend between 1996 and 25 (ignoring 1995 due to low data capture) corresponding to the reduction in background NO X and NO 2. AEA Energy & Environment 8

20 2.2.5 Germany Figure 2.1e shows that for roadside sites in Baden Wűrttemberg, there was a significant decrease in NO X concentrations between 1995 and 25. This decrease is not reflected in NO 2 trends at all the roadside sites as measured annual mean NO 2 concentrations went up at STS (Stuttgart Strassenstation) over this period and only decreased slightly at KAS (Karlsruhe Strassenstation). At background sites, annual mean NO X concentrations decreased marginally between 1995 and 25 and annual mean NO 2 concentrations were fairly constant. In terms of exceedences of the annual mean limit value for NO 2, all the roadside sites selected for investigation exceeded 4μg m -3, with a maximum annual mean concentration of 8.3μg m -3 at STS in 23. Only one background site, STZ (Stuttgart Zuffenhausen) has exceeded the limit value since 1999 with annual mean NO 2 concentrations ranging between 4 and 5μg m -3 since Ozone, as measured at the background sites, shows a gradual increase between 1995 and 25, with clear evidence of a peak in Greece Trends in Annual mean concentrations of NO X, NO 2 and ozone in the Athens case study (Figure 2.1f ) are slightly less apparent than for some of the other case studies because several of the points plotted have low data capture (<75%). These points have been included in our analyses despite this low data capture because only a minimum of 1 hours of data in a year are necessary to run the Netcen Primary NO 2 model. A low data capture therefore does not necessarily prevent the model from running so we need to understand the input monitoring data even where the data capture is low. Generally, Figure 2.1f suggests NO X concentrations in Athens are coming down, while this trend is not so apparent in the NO 2 data. All four roadside monitoring sites in Athens exceeded the annual mean NO 2 limit value in 24, with a maximum concentration of 88.3μg m -3 at GR32A (PATISION). For both NOx and NO 2, the roadside site GR22A (MAROUSI) and the background site GR35A (LYKOVRISI) have very similar annual mean concentrations. In terms of ozone, these two sites also exhibited similar behaviour with a sharp decrease in concentrations between 2 and 21 and then constant concentrations since then. The roadside site GR2A (ATHINAS) showed a very different distribution of ozone concentrations between 2 and 24 with sharp increases and decreases apparent Italy Annual mean NO X concentrations in the Milan case study (Figure 2.1g) roadside sites show a significant downward trend between 2 and 24. NO 2 concentrations show less of an overall trend with concentrations at some sites decreasing (e.g. IT75A (VERZIERE 3154)) and some increasing (e.g. IT477A (MARCHE 31526)). All four roadside sites selected had annual mean NO 2 concentrations significantly in excess of the 21 annual mean limit value. At the two background sites considered, there was a significant difference in terms of both NO X and NO 2 concentrations. IT117A (P.CO LAMBRO 3153), which is located centrally in Milan, had very high concentrations with the annual mean NO 2 limit value exceeded across all years considered. By contrast, IT123A (ARCONATE 3154) had much lower NO X and NO 2 concentrations reflecting its location slightly outside of Milan. Ozone at both these sites peaked in 23, with concentrations at IT123A staying high in The Netherlands In the Netherlands (Figure 2.1h), roadside sites for annual mean NO X concentrations exhibit a slight downward trend between 2 and 24 with the exception of NL248A (Breukelen-Snelweg) where NO X concentrations dipped in before returning to a similar level to 2 in 24. At background sites, there was almost no change in NO X concentrations across these five years. In terms of NO 2 concentrations, the roadside sites selected show a slight increase between 2 and 24, with all sites exceeding the limit value of 4μg m -3. The highest exceedence in 24 of 61.μg m -3 occurred at the relatively new monitoring site NL253A (Den Haag-Veerkade). There is currently little indication of the likely trend in NO 2 concentrations at this site since the annual mean concentration for AEA Energy & Environment 9

21 23 only had 26% data capture. Background NO 2 concentrations remained relatively constant across the period of investigation with a slight peak in concentrations in 23. Ozone at the roadside sites increased across this period, as did ozone at the background sites although the trend at the background sites is less pronounced Spain Three of the four roadside sites in the Barcelona case study (Figure 2.1i) generally showed downward trends in NO X concentrations over the six years considered, although there was a slight increase in 25. Annual mean NO 2 concentrations at two of the roadside sites, ES148A (Gracia Sant Gervasi) and ES1438A (Barcelona), increased over this period, while at the two remaining roadside sites, NO 2 concentrations remained more level. Concentrations of NO X and NO 2 at the background site remained relatively constant. The trends in ozone concentration were generally upwards at the roadside sites but downward at the background site UK In London (Figure 2.1j) annual mean NO X and NO 2 concentrations at the selected sites fell into three distinct groups, listed from lowest to highest: background sites; roadside sites excluding London Marylebone Road (GB682A); and London Marylebone Road. NO X concentrations at all these sites have fallen between the start of monitoring and 25. In terms of NO 2 concentrations, the distinction between different types of sites was less clear than for NO X. All but one of the background sites along with all the roadside sites exceeded the annual mean NO 2 limit value in 24 and most earlier years. It is difficult to pick out any definite upward or downward trend in the NO 2 data for most of the sites. However, NO 2 concentrations do appear to have come down at London Cromwell Road 2 (GB695A) and, since 23, to be increasing at Camden Kerbside (GB636A). At London Marylebone Road there is a clear step change in annual mean NO 2 concentrations in 23 and 24 which may reflect the introduction of a bus lane next to the site and the introduction of more buses with particulate filters attached (AQEG, 26). Ozone concentrations at UK sites showed a general upward trend over the period of interest. Background annual mean concentrations of ozone were significantly higher than at London Marylebone Road. AEA Energy & Environment 1

22 Table 2.2.a Details of roadside monitoring site included analysis presented in this report Country EOI Code Site Name Latitude Longitude Date of 1 st measurement NO 2 /NO X O 3 Data source Paired background site Austria AT168A Salzburg Mirabellplatz '21'' +13 2'45'' Airbase (26) AT75A Austria AT21A Zederhaus +47 9'15'' +13 3'18'' Airbase (26) AT75A Austria AT38A Salzburg Rudolfspaltz '51'' +13 3'13'' Airbase (26) AT75A Austria AT24A Hallein Hagerkreuzung ''' +13 6'6'' Airbase (26) AT75A Czech Republic CZ66A Pha2-Legerova +5 4'21'' '48'' Airbase (26) CZ2A Czech Republic CZ8A Pha1-nam. Republiky +5 5'18'' '46'' Airbase (26) CZ2A Czech Republic CZ13A Pha1-Vrsovice +5 4'2'' '51'' Airbase (26) CZ2A Czech Republic CZ11A Pha5-Mlynarka +5 4'2'' '12'' Airbase (26) CZ2A Czech Republic CZ65A Pha5-Smichov +5 4'24'' '57'' Airbase (26) CZ2A Finland FI16A Tikkurila '24'' +25 2'22'' Airbase (26) FI124A Finland FI4A Vallila '37'' '5'' Airbase (26) FI124A Finland FI6A Vesku '2'' '44'' Airbase (26) FI124A Finland FI16A Turun kauppatori +6 27'4'' '3'' Airbase (26) FI124A France FR895A Boulevard périphérique Auteuil ''' +2 15'12'' Airparif (26) FR918A France FR335A Place Victor Basch '4'' +2 19'39'' Airparif (26) FR918A France FR898A Autoroute A1 - Saint-Denis '39'' +2 21'26'' Airparif (26) FR918A France FR91A Quai des Célestins '1'' +2 21'38'' Airparif (26) FR918A Germany MAV Manheim Strassenstation '37'' +8 28'19'' LUBW (26) MAN and MAS Germany STS Stuttgart Strassenstation '5'' +9 1'49'' LUBW (26) STB and STS Germany KAS Karlsruhe Strassenstation +49 '32'' +8 23'16'' LUBW (26) KAN Germany FNV Freiburg Strassenstation '57'' +7 51'13'' LUBW (26) FMI Greece GR22A MAROUSI +38 1'51'' '17'' Airbase (26) GR35A Greece GR4A GOUDI '4'' '4'' Airbase (26) GR35A Greece GR32A PATISION '57'' '59'' Airbase (26) GR35A Greece GR2A ATHINAS '41'' '37'' Airbase (26) GR35A Italy IT477A MARCHE '38'' +9 12'27'' Airbase (26) IT117A, IT123A Italy IT771A COMO CENTRO '16'' +9 5'1'' Airbase (26) IT117A, IT123A Italy IT116A SENATO MARINA '11'' +9 11'53'' Airbase (26) IT117A, IT123A Italy IT75A VERZIERE '46'' +9 11'45'' Airbase (26) IT117A, IT123A Netherlands NL248A Breukelen-Snelweg '11'' +4 59'19'' Airbase (26) NL229A AEA Energy & Environment 11

23 Country EOI Code Site Name Latitude Longitude Date of 1 st measurement NO 2 /NO X O 3 Data source Paired background site Netherlands NL253A Den Haag-Veerkade +52 4'3'' +4 18'58'' Airbase (26) NL199A Netherlands NL234A Utrecht-Erzeijstraat +52 4'8'' +5 7'16'' Airbase (26) NL229A Netherlands NL224A Vlaardingen-Floreslaan '41'' +4 19'37'' Airbase (26) NL199A Netherlands NL244A Haarlem-Amsterdamsevaart '54'' +4 38'53'' Airbase (26) NL25A Netherlands NL235A Utrecht-de Jongweg +52 6'22'' +5 7'31'' Airbase (26) NL229A Departament de Medi Spain ES1453A ES1453A-II-TORREBALLDOVINA '12'' +2 12'36'' Ambient i Habitatge (Generalitat de Catalunya) Sant Culgat del Vallès (26) Departament de Medi Spain ES1262A ES1262A-AD-SABADELL '46'' +2 6'1'' Ambient i Habitatge (Generalitat de Catalunya) Sant Culgat del Vallès (26) Departament de Medi Spain ES148A ES148A-IJ-GRACIA-SANT Ambient i Habitatge '1'' +2 9'16'' GERVASI (Generalitat de Catalunya) (26) Sant Culgat del Vallès Spain ES1438A ES1438A-IH- BARCELONA(Eixample) '1'' +2 9'19'' Departament de Medi Ambient i Habitatge (Generalitat de Catalunya) (26) Sant Culgat del Vallès UK GB636A Camden Kerbside '41'' - 1'31'' UK Air Quality Archive (26) London N. Kensington UK GB659A London A3 Roadside '25'' - 17'31'' UK Air Quality Archive (26) London Teddington UK GB695A London Cromwell Road '44'' - 1'43'' UK Air Quality Archive (26) London N. Kensington UK GB682A London Marylebone Road '21'' - 9'17'' UK Air Quality Archive (26) London N. Kensington UK GB667A Southwark Roadside '55'' - 3'46'' UK Air Quality Archive (26) London Southwark UK GB624A Tower Hamlets Roadside '22'' - 2'32'' UK Air Quality Archive (26) London Hackney AEA Energy & Environment 12

24 Table 2.2.b Details of background monitoring site included analysis presented in this report Country EOI Code Site Name Latitude Longitude Date of 1 st measurement NO 2 /NO X O 3 Data source Austria AT75A Salzburg Lehen '3'' +13 1'48'' Airbase (26) Czech Republic CZ2A Pha4-Libus +5 '27'' '56'' Airbase (26) Finland FI124A Kallio '15'' '2'' Airbase (26) France FR918A PARIS 6ème +48 5'56'' +2 2'9'' Airparif (26) Germany MAN Mannheim Nord '42'' +8 28'1'' LUBW (26) Germany MAS Mannheim Sud '6'' +8 31'35'' LUBW (26) Germany STB Stuttgart Bad Cannstatt '35'' +9 13'51'' LUBW (26) Germany STZ Stuttgart Zuffenhausen '36'' +9 1'25'' LUBW (26) Germany KAN Karlsruhe Nordwest +49 1'47'' +8 21'23'' LUBW (26) Germany FMI Freiburg Mitte +48 '9'' +7 49'57'' LUBW (26) Greece GR35A LYKOVRISI +38 4'12'' '37'' Airbase (26) Italy IT117A P.CO LAMBRO '56'' +9 14'51'' Airbase (26) Italy IT123A ARCONATE '48'' +8 5'55'' Airbase (26) Netherlands NL199A Den Haag-Rebecquestraat +52 4'41'' +4 17'21'' Airbase (26) Netherlands NL229A Zegveld-Oude Meije +52 8'2'' +4 5'18'' Airbase (26) Netherlands NL25A De Zilk-Vogelaarsdreef '53'' +4 3'37'' Airbase (26) Spain Sant Culgat del Vallès '24 Departament de Medi Ambient i Habitatge (Generalitat de Catalunya) (26) UK GB62A London N. Kensington '16'' - 12'48'' UK Air Quality Archive (26) UK GB644A London Teddington '16'' - 2'23'' UK Air Quality Archive (26) UK GB656A London Southwark '26'' - 5'48'' UK Air Quality Archive (26) UK GB65A London Hackney '32'' - 3'24'' UK Air Quality Archive (26). AEA Energy & Environment 13

25 Figure 2.1. Recent trend in NO X, NO 2 and Ozone (where available) for sites listed in Table 2.2. All points have greater than 75% data capture unless otherwise indicated. a) Austria b) Czech Republic Annual mean concentration (ugm-3) NOx AT168A Roadside NOx AT21A Roadside NOx AT38A Roadside NOx AT24A Roadside NOx AT75A background Annual mean concentration (ugm-3) NOx CZ66A Roadside NOx CZ8A Roadside NOx CZ13A Roadside NOx CZ11A Roadside NOx CZ65A Roadside NOx CZ2A background Data Capture 42% Annual mean concentration (ugm-3) NO2 AT168A Roadside NO2 AT21A Roadside NO2 AT38A Roadside NO2 AT24A Roadside NO2 AT75A background Annual mean concentration (ugm-3) NO2 CZ66A Roadside NO2 CZ8A Roadside NO2 CZ13A Roadside NO2 CZ11A Roadside NO2 CZ65A Roadside NO2 CZ2A background Data Capture 35% O3 CZ8A Roadside O3 CZ65A Roadside Annual mean concentration (ugm-3) O3 AT168A Roadside O3 AT75A background Annual mean concentration (ugm-3) O3 CZ2A background Data Capture 7% AEA Energy & Environment 14

26 Figure 2.1. continued c) Finland d) France Annual mean concentration (ugm-3) NOx FI16A Roadside NOx FI4A Roadside NOx FI6A Roadside NOx FI16A Roadside NOx FI124A background Annual mean concentration (ugm-3) Data Capture 28% Data Capture 69% NOx FR895A Roadside NOx FR335A Roadside NOx FR898A Roadside NOx FR91A Roadside NOx FR918A background Data Capture 59% Data Capture 64% Annual mean concentration (ugm-3) NO2 FI16A Roadside NO2 FI4A Roadside NO2 FI6A Roadside NO2 FI16A Roadside NO2 FI124A background Annual mean concentration (ugm-3) Data Capture 7% Data Capture 36% 4 NO2 FR895A Roadside NO2 FR335A Roadside 2 NO2 FR898A Roadside NO2 FR91A Roadside NO2 FR918A background AEA Energy & Environment 15

27 6 45 O3 FI124A background 4 O3 FR918A background 5 Annual mean concentration (ugm-3) Annual mean concentration (ugm-3) Data Capture 36% Figure 2.1. continued e) Germany f) Greece Annual mean concentration (ugm-3) NOx FNV Roadside NOx KAS Roadside NOx MAV Roadside NOx STS Roadside NOx FMI background NOx KA N background NOx MA N background NOx STB background NOx MAS background NOx STZ background Annual mean concentration (ugm-3) NOx GR22A Roadside NOx GR4A Roadside NOx GR32A Roadside NOx GR2A Roadside NOx GR35A background Data Capture 72% Data Capture 61% Data Capture 66% Data Capture 39% Data Capture 65% AEA Energy & Environment 16

28 Annual mean concentration (ugm-3) NO2 FNV Roadside NO2 KAS Roadside 1 NO2 MAV Roadside NO2 STS Roadside NO2 FMI background NO2 KAN background NO2 MAN background 1995 NO2 1996STB background NO2 MAS background NO2 STZ background Annual mean concentration (ugm-3) NO2 GR22A Roadside NO2 GR4A Roadside NO2 GR32A Roadside NO2 GR2A Roadside NO2 GR35A background Data Capture 72% Data Capture 61% Data Capture 7% Data Capture 66% Data Capture 39% Data Capture 65% Annual mean concentration (ugm-3) O3 FMI background 2 O3 KAN background O3 MAN background O3 STB background 1 O3 MAS background O3 STZ background Annual mean concentration (ugm-3) 8 O3 GR22A Roadside 7 O3 GR32A Roadside O3 GR2A Roadside 6 O3 GR35A background Data Capture 47% Data Capture 7% Figure 2.1. continued g) Italy h) Netherlands AEA Energy & Environment 17

29 Annual mean concentration (ugm-3) NOx IT477A Roadside NOx IT771A Roadside NOx IT116A Roadside NOx IT75A Roadside NOx IT117A background NOx IT123A background Annual mean concentration (ugm-3) NOx NL248A Roadside NOx NL253A Roadside NOx NL234A Roadside NOx NL224A Roadside NOx NL244A Roadside NOx NL235A Roadside NOx NL199A background NOx NL229A background NOx NL25A background Data Capture for NL229A 74% Data Capture for NL25A 61% Data Capture 26% Annual mean concentration (ugm-3) NO2 IT477A Roadside NO2 IT771A Roadside NO2 IT116A Roadside NO2 IT75A Roadside NO2 IT117A background NO2 IT123A background Annual mean concentration (ugm-3) NO2 NL248A Roadside NO2 NL253A Roadside NO2 NL234A Roadside NO2 NL224A Roadside NO2 NL244A Roadside NO2 NL235A Roadside NO2 NL199A background NO2 NL229A background NO2 NL25A background Data Capture 26% Annual mean concentration (ugm-3) O3 IT117A background O3 IT123A background Annual mean concentration (ugm-3) O3 NL248A Roadside O3 NL234A Roadside O3 NL224A Roadside O3 NL235A Roadside O3 NL199A background O3 NL229A background O3 NL25A background Data Capture 73% Figure 2.1. continued i) Spain AEA Energy & Environment 18

30 j) UK Annual mean concentration (ugm-3) Data Capture 59% Data Capture 69% NOx ES1453A Roadside NOx ES1262A Roadside NOx ES148A Roadside NOx ES1438A Roadside NOx Sant Cugat del Vallès background Annual mean concentration (ugm-3) NOx Camden Kerbside NOx London A3 Roadside NOx London Cromwell Road 2 NOx London Marylebone Road NOx Southwark Roadside NOx Tower Hamlets Roadside NOx London N. Kensington NOx London Teddington NOx London Southwark NOx London hackney Annual mean concentration (ugm-3) Data Capture 59% NO2 ES1453A Roadside NO2 ES1262A Roadside NO2 ES148A Roadside NO2 ES1438A Roadside NO2 Sant Cugat del Vallès background Data Capture 71% Annual mean concentration (ugm-3) NO2 Camden Kerbside 2NO2 London A3 Roadside NO2 London Cromwell Road 2 NO2 London Marylebone Road NO2 Southwark Roadside NO2 Tower Hamlets Roadside NO2 London N. Kensington Background 1995 NO2 London 1996 Teddington NO2 London Southwark NO2 London hackney Annual mean concentration (ugm-3) Data Capture 71% 1 O3 ES1453A Roadside O3 ES1262A Roadside O3 ES148A Roadside 5 O3 ES1438A Roadside O3 Sant Cugat del Vallès background UK points with low data capture: O3 Camden Kerbside % NO 2 and NO X Camden Kerbside %, 22 8%, 23 44%, 24 39% Annual mean concentration (ugm-3) O3 London Marylebone Road O3 london n. kensington O3 london teddington O3 London Southwark O3 london hackney AEA Energy & Environment 19

31 NO 2 and NO X London A3 Roadside % NO 2 and NO X London Cromwell Road % NO 2 and NO X London Marylebone Road % NO 2 and NO X Southwark Roadside % % % NO 2 and NO X Tower Hamlets Roadside % NO 2 and NO X London North Kensington % NO 2 and NO X London Teddington % NO 2 and NO X London Southwark % 23 73% 2.3 Summary The data presented in Figure 2.1 shows that across most case study areas, while NO X concentrations are generally levelling off or starting to decline, NO 2 concentrations are generally remaining high, and in some cases continuing to increase. This suggests that there is some other influencing factor existing at many of these sites to keep NO 2 concentrations disproportionately high. However, because of inter-site differences, several sites in the same case study area can behave slightly differently and the behaviour at each site is best understood in terms of local factors, especially the characteristics of the road on which the site is located (e.g. London A3 roadside behaved very differently to London Marylebone Road because they are located on very different types of roads with different traffic flows and compositions). AEA Energy & Environment 2

32 3 Recent Trends in f-no 2 using the Netcen Primary NO 2 model 3.1 Introduction f-no 2 is very difficult to measure directly and few measurement studies are available. Therefore, other approaches are necessary to analyse recent trends in f-no 2. Two approaches taken in this study include an emissions inventory approach (see section 4) and a modelling approach based on ambient monitoring data. The second of these is presented in this section. The model selected for use here is the Netcen Primary NO 2 model. Box 1 (see section 1.3) gives a brief introduction to this. A more detailed description of the model and of its application to monitoring sites across the UK is given in Abbott (26). Module 1 of the model, the analysis module, calculates an annual estimate of f-no 2 for a given roadside location using regression analysis on hourly modelled f-no 2 values. Input data required by the model include hourly measured NO X, NO 2 and ozone at the roadside site and at a background site selected to represent the non-roadside concentrations at this roadside site. Limiting the selection of roadside monitoring sites to those that measure ozone however would severely limit this analysis as many of the worst sites in terms of exceedences of the annual mean NO 2 limit value do not monitor ozone and some of the case study areas actually had no roadside ozone monitoring at all. Therefore, for roadside sites with no ozone monitoring, ozone has been modelled using module 2: the ozone module. Where roadside ozone measurements are available, the analysis module of the model has been run twice, once using the roadside ozone data and once using the ozone module to check how well the ozone module replicates actual measured ozone at these sites. Table 2.2a in the previous section showed which background site is paired to each roadside site. Table 3.1 below, presents details of which meteorological station has been used to run the ozone module for each case study. Where met data has been used that has a different time zone to the NO X, NO 2 and ozone monitoring data, the met data has been re-aligned to match the monitoring data s time zone. Table 3.1. Details of meteorological stations used for each case study Country City (if applicable) Met station used Code Latitude Longitude Austria Salzberg/Hallein Milan Linate IML 45.4N 9.3E Czech Republic Prague Laupheim THL 48.2N 9.9E Finland - Amsterdam HAM 52.3N 4.8E Warsaw Okecie 1 PWA 52.1N 21.E France Paris Paris Orly FPO 48.7N 2.4E Germany Baden Wűrttemberg Laupheim THL 48.2N 9.9E Greece Athens Athens 2 ATH Italy Milan Milan Linate IML 45.4N 9.3E Netherlands - Amsterdam HAM 52.3N 4.8E Spain Barcelona Barcelona Airport EBL 41.2N 2.1E UK London London Heathrow HEAT 51.4N.5W 1 For Finland, two model runs were carried out to assess the sensitivity of the modelling to using different met data 2 In Athens, there were no met stations with complete data for the entire 6 year period of interest. Therefore, 3 data sets were combined. These were: Elefsis (38.67 N, E), Athens/Hellenkion (37.9 N, E) and Eleftherios ( N, E) When running the analysis module, the model loops through each hour of the year and then calculates the annual mean f-no 2 through regression analysis of hourly f-no 2 for all the hours modelled. If any of the model input data is missing for any hour that hour will not be modelled. Similarly, if the modelled f- NO 2 >1 for any given hour, this hour is not included in the regression analysis. This is because f-no 2 >1 is not physically realistic because the maximum possible proportion of NO X that can be NO 2 is 1%, (i.e. f-no 2 = 1). Where f-no 2 >1 occurs, it suggests that the model assumptions are not holding true. For example, in Greece, we found a modelled f-no 2 >1 for a significant number of hours. In AEA Energy & Environment 21

33 these hours, it is likely, given the ozone climate of Greece that reactions occurred between the background and roadside site that added extra total oxidant to the air mass. It is also possible for any given hour, that the air mass at the background site does not represent well the behaviour of the nontraffic related component of the roadside site. The analysis module performs better where there is a significant NO X roadside increment. Therefore, a minimum roadside increment was set at 1μg m -3. For hours with a roadside increment less than this, the model did not run and no result was included in the regression analysis to calculate the annual f- NO Ozone Module Verification For roadside sites where ozone measurements were made over the period of interest for the case studies, it has been possible to test the ozone module performance by comparing modelled hourly ozone concentrations with measured hourly ozone concentrations. This verification is important to carry out because the EU covers a range of ozone climates, and model performance may vary between these climates. Table 3.2 shows the R 2 values (calculated as the square of the Pearson product moment correlation co-efficient) for measured versus modelled hourly ozone concentrations at all roadside monitoring sites where ozone measurements were available between 2 and 25. This shows that at sites in Austria, the Czech Republic, the Netherlands and the UK the R 2 value values are typically sufficiently high (e.g. >.75) to suggest that the model is performing very well and most of the variability in the measured data can be explained by the modelled data. For sites in Greece and Spain, the model generally performs less well than at sites in the more northerly member states. However, most of the R 2 values were greater than.5 in these countries. This means that over 5% of the variability in the modelled ozone can be explained by the measured ozone and therefore the model is performing sufficiently well to use it in these member states. Table 3.2. R 2 values for measured and modelled ozone at roadside sites with ozone monitoring R 2 Value Country Site Austria AT168A Czech Republic CZ8A CZ65A GR22A Greece GR32A.52 GR2A NL248A Netherlands NL234A NL224A NL235A ES1453A Spain ES1262A ES148A ES1438A UK London Marylebone Road Scatter plots of measured versus modelled ozone and time series plots of measured and modelled ozone for a typical site from each case study are presented in Appendix Best Estimate f-no 2 Results Time series plots of modelled annual f-no 2 for each case study are presented in Figure 3.1. For each case study, two graphs are shown. The graph on the left presents the modelled annual f-no 2 where AEA Energy & Environment 22

34 more than 1 hours have been modelled per year (i.e. >.1% of total hours per year have been modelled). The graph on the right applies a more stringent threshold, so the f-no 2 is only shown where more than 3% of hours have been successfully modelled per year. This threshold may seem low compared with data capture thresholds set in EU law for the reporting of monitoring data to the European Commission in the questionnaire. However, the complex nature of the model which relies upon having six channels of ambient monitoring data (three at the roadside site and three at it s paired background site) and meteorological data if required, along with a minimum NO X roadside increment of 1μg m -3 and the maximum hourly f-no2 limit of one for each hour, means that many hours are not modelled. For example, if all other criteria for the model to run were met for a given hour except that there was gap in the background site ozone measurements for that hour, the model would not run for that hour. This means that a threshold of 3% of hours modelled is actually relatively stringent. The decision of where exactly to set this minimum threshold of 3% of hours modelled per year is based upon an analysis of primary oxidant from traffic (calculated as f-no 2 multiplied by the NO X roadside increment) and secondary oxidant associated with non road transport sources (this is the intercept of the regression of hourly f-no 2 used in the model to calculate the annual f-no 2 ). Appendix 3 contains graphs for each case study showing how the model estimate of these two sources of oxidant (ppb) vary at each of the monitoring sites considered in this analysis. In general, at sites where there is a lot of primary oxidant from traffic, the model works well and at sites where there is less primary oxidant, the model works less well. Where there are significant levels of secondary oxidant from background sources, this can lead to the signal from the traffic related oxidant being much less certain. Table 3.3. presents tabulated modelled f-no 2 results for years with greater than 3% of hours modelled. The percentage of hours modelled per year for each model run is also presented here, but f-no 2 is only given where this percentage is greater than 3% Austria In Salzberg and Hallein (Figure 3.1a), there were two sites with more than 3% of hours modelled per year: AT38A (Salzburg Rudolfspaltz) and AT24A (Hallein Hagerkreuzung). f-no 2 at AT38A increased from 7.2% in 2 to 12.3% in 24. The trend at AT24A was generally flatter with only a 1.3% increase between 2 and 24. At other sites selected in Austria, the model data capture was less than 3%. This probably reflects the low NO X roadside increment at these sites (see Figure 2.1a) Czech Republic For all sites modelled in Prague, with the exception of the sensitivity test on site CZ65A _1, Pha5- Smichov, (Figure 3.1b) there were at least two years with greater than 3% of hours modelled. For run CZ66A_1 (Pha2-Legerova), the highest f-no 2 was recorded: 2.2% in 24. The trend at this site might be misleading because the availability of monitoring data in 23 was heavily weighted towards the second half of the year. This trend will become clearer as future years data becomes available. At two of the other sites, CZ8A (Pha1-nam. Republiky) and CZ65A (Pha5-Smichov), there was a slight upward trend in modelled f-no 2 between 2 and 24, while at CZ11A (Pha5-Mlynarka) the overall trend was flatter and at CZ13A (Pha1-Vrsovice) there was a slight downward trend Finland For the Finish case study area (Figure 3.1c), very few of the model runs had greater than 3% data capture. This probably reflects the low levels of primary oxidant associated with road transport at these sites (see Appendix 3). However the model results do suggest that f-no 2 is not a problem at the Finish sites considered since none of the model runs without a data capture threshold applied show any upward trend in f-no 2 or high absolute values of f-no 2. Also, since NO X and NO 2 concentrations at the Finish sites are relatively low compared with other sites considered in this study (see Figure 2.1c), even if there were high f-no 2 values associated with these sites, this would probably be insufficient to push ambient NO 2 concentrations over the 21 annual mean limit value. AEA Energy & Environment 23

35 3.3.4 France At the sites modelled in Paris (Figure 3.1d), the threshold of 3% of hours modelled per year was met at all sites for all years from 1996 onwards. Generally, the data capture was significantly higher than this threshold suggesting the model results are very reliable. Analysis of the primary and secondary oxidants at the sites also suggests the model should work well since there were typically high primary traffic related oxidant concentrations and low secondary non-traffic oxidants (see Appendix 3). Overall, f-no 2 at the sites selected in Paris increased from between 3.2 and 7.5% in 1996 to between 9.4% and 13.8% in 25. This suggests that although in France the penetration of diesel vehicles into the fleet may have started early, f-no 2 was not significantly higher in the period from 1995 to 2 than at the case study areas in Germany and the UK for which we have estimated f-no 2 for these years Germany In the Baden Wűrttemberg case study (Figure 3.1e) high level of data capture was achieved. However, in contrast to the situation in Paris, the rise in f-no 2 has been much steeper. For the first five years considered, , there was no clear overall trend in f-no 2. Since 1999, there has been a steady increase in f-no 2 from % in 1999 to % in 25. This broadly agrees with the findings of Kessler et al (26), who used the model presented by Carslaw and Beavers (26) at the same combination of sites that have been used here. The sharp drop in f-no 2 at the roadside site in Manheim (MAV Manheim Strassenstation) may be explainable by the presence of construction work near the site from mid 24 until late 25, which effected traffic flow along the road this site is located on (Werner Scholz pers. Comm. 27) Greece Model runs for the Athens case study (Figure 3.1f) generally had poor overall data capture. This was due to a variety of factors including gaps in the NO X, NO 2 and ozone input data and also the relatively high secondary oxidant levels and low primary oxidant at the sites selected (see Appendix 3). For the sites with greater than 3% of hours modelled, there seems to be a slight upward trend in f-no 2. However, to draw firm conclusions regarding this trend, it would be necessary to have more complete model results Italy For Italy (Figure 3.1g), two sets of model runs were carried out because the background site IT117A (P.CO LAMBRO 3153), which is situated in the centre of Milan had a relatively high NO X concentration over the years considered for a background site (see Figure 21.g). This might be representative of concentrations in background locations in central Milan. However, it is also possible that this site is influenced by nearby roads. Therefore, it was decided to run the Netcen Primary NO 2 model to calculate recent trends in f-no 2 using this site as the background site, but also to do sensitivity analysis on the impact on modelled f-no 2 of using a more rural site, IT123A (ARCONATE 3154). The model results using IT117A are presented here and analysis of the impact of using the more rural background site IT123A is presented in section 3.4 on sensitivity analysis. Generally, there seems to be very little evidence of a consistent trend across the case study area in Milan in terms of changes in modelled f-no 2. This may reflect the relatively poor performance of the model in this location where local photochemistry may be more important than in north west Europe Netherlands Sites in the Netherlands case study (Figure 3.1h) with greater than 3% of hours modelled per year showed a general upward trend between 2 and 24 with f-no 2 in 2 lying in the range of % compared with % in 24. AEA Energy & Environment 24

36 3.3.9 Spain Modelled f-no 2 trends at the sites in Barcelona (Figure 3.1i) showed a very mixed picture. At ES1453A (ES1453A-II-TORREBALLDOVINA) modelled f-no 2 fell from 23.8% in 2 to 13.6% in 24. At all other sites, there was a modelled increase in f-no 2 between 2 ( %) to 24 ( %). The model results therefore suggest that f-no 2 at roadside sites Barcelona has been generally relatively high compared with sites in other case study areas UK In the London case study (Figure 3.1j), f-no 2 at all sites increased between 1996 and 25, although the trend at most sites started to level off after 23. London Cromwell Road (GB695A), when it opened in 1998 had an f-no 2 of 15.6%, which was comparatively high for Since then, f-no 2 at this site has increased to 22.8% in 25. The only other site in the London case study to exceed f-no 2 of 2% was London Marylebone Road (GB682A), which had an f-no 2 of 21.6% in 25. At this site, there was a step change in f-no 2 between 22 and 23 where f-no 2 increased from 1.1% to 19.1%. This is probably attributable to the introduction of a bus lane next to the site in 22 combined with the introduction of more buses with particulate filters in London (AQEG, 26). Other noteworthy sites in the London case study include London A3 Roadside (GB659A). This site had a very low modelled f-no 2 in 1997 when it opened, then between 1999 and 23, there was a steep increase in f-no 2 from.6 to 12.8%, before levelling off after 23. AQEG (26) have modelled f-no 2 using the model of Carslaw and Beavers (26) for several of the roadside sites considered here. For comparable sites, the results show good agreement in terms of trends, but the model of Carslaw and Beavers (26) generally predicts slightly higher f-no 2 than the Netcen Primary NO 2 model. 3.4 f-no 2 Sensitivity Analysis The graphs in Figure 3.1. show both our best estimate of f-no 2 at each site (black lines) and model results carried out as sensitivity analysis (red lines). Three sets of sensitivity tests have been done as follows: Sensitivity of modelled f-no 2 to the ozone module (all sites where roadside ozone measurements were available). Sensitivity of modelled f-no 2 to met data (Finish sites only). Sensitivity of modelled f-no 2 to using different paired background sites (German and Italian sites) The sensitivity of modelled f-no 2 to using the ozone module was tested at a range of sites. At CZ8A (Pha1-nam Republicky), model calculations using the ozone module over-predicted f-no 2 compared to using roadside ozone measurements by between 1 and 3%. However, the overall trends in f-no 2 for the two model runs were very similar. In the Athens case study, modelled f-no 2 using the roadside ozone module closely reflected modelled f-no 2 with measured roadside ozone data. However, because of gaps in the ozone monitoring data at some of the roadside sites in Athens, the data capture using the ozone module was generally substantially better than when using measured ozone. In the Netherlands, the ozone module performed very well, closely following the f-no 2 predictions with measured roadside ozone input data. In the Barcelona case study, the modelled f-no 2 using measured and modelled roadside ozone do not show particularly close agreement in terms of absolute values, but the trends follow fairly closely. As with Greece, the number of hours modelled was significantly higher using the ozone module than using measured ozone in Barcelona. At London Marylebone Road (GB682A) in London, the sensitivity test showed very close agreement with the modelled f-no 2 using measured roadside ozone. Implications of this sensitivity analysis include the following. For analysis of recent trends in f-no 2 at sites without roadside ozone measurements using the Netcen Primary NO 2 model, it shows that in most locations, it is likely that the modelled f-no 2 would not be significantly different if t ozone measurements were available. This is suggested by the fact that in the Czech Republic, the Netherlands and the UK there was close agreement between our best estimate and the sensitivity test. These case study areas are broadly representative in terms of ozone climate of most of the other AEA Energy & Environment 25

37 case study locations. Therefore if the ozone module performs well in these locations, it is likely it will work well in other case study areas too. The exception to this comes in the southern parts of the EU where both in Spain and Greece, the f-no 2 modelled using the ozone module mirrored f-no 2 modelled with roadside ozone measurements less closely. This probably reflects the lower R 2 values for these locations between measured and modelled ozone in these case studies as discussed in section 3.2 above. However, using the ozone module did enable the f-no 2 model to run for a greater number of hours in these areas than otherwise would be the case. In terms of future projections of ambient NO 2 concentrations, this sensitivity analysis also has implications because a variant of the ozone module is used at all sites to calculate projected roadside ozone concentrations in the future. These modelled ozone concentrations then feed into the model projections of future ambient NO 2 concentrations. The close agreement between the f-no 2 using measured roadside ozone and the ozone module suggests that using the ozone module in this way will not have a significant detrimental impact on the modelled future ambient concentrations. The second set of sensitivity analysis, testing modelled f-no 2 values response to the use of different sets of input data was carried out in Finland. The reason for selecting Finland for this analysis was that for all the other case studies, nearby met data was available whether from within that case study area, or from a neighbouring country. By contrast, for Finland, we have had to use met data from further away. Therefore two sets of met data, Amsterdam Airport and Warsaw Okecie, were used. Figure 3.1c and Table 3.2 show that, for this case study, using different met data had little to no difference on the modelled f-no 2 results. This strongly suggests that had met data from within Finland been available, it would not have significantly impacted on the modelled f-no 2 values in Finland. The third set of sensitivity analysis, testing the modelled f-no 2 response to use of different paired background sites was carried out in Germany and Italy. In Germany, two different background sites in Stuttgart were paired with the roadside site STS (Stuttgart Strassenstation) and in Manheim two different background sites were paired with MAV (Manheim Strassenstation). Figure 3.1e shows that using these two different background sites had very little impact upon the modelled f-no 2 for these roadside sites. By comparison, Kessler et al (26) using the model of Carslaw and Beevers (25) found that f-no 2 for weekdays (i.e. Monday to Friday) for the identical set of sites was sometimes sensitive to which background site was paired with the roadside site. Their results show that the magnitude of the effect of using the alternative background sites on f-no 2 varied significantly between years. In Italy, the use of different background sites had a greater impact (Figure 3.1g) than in the German case study. This is because IT117A (P.CO LAMBRO 3153), despite being classified as a background site, had high ambient NO X and NO 2 concentrations relative to other background sites considered in this study (see Figure 2.1g). These high concentrations can be attributed to the site s location in the centre of Milan and effectively lowers the roadside increment for the roadside sites with which it is paired, causing fewer hours to be modelled using this background site. This makes the f- NO 2 model results relatively variable between years because of the low levels of modelled hourly data available. In contrast IT123A (ARCONATE 3154) is located outside Milan and is classified as rural. Therefore NO X and NO 2 concentrations are significantly lower at this site, so the NO X roadside increment is higher and more hours are modelled. This results in more stable trends in modelled f-no 2 than from using the urban background site. AEA Energy & Environment 26

38 Figure 3.1. Recent modelled trends in f-no 2. Lines plotted in black are our best estimate; lines in red are sensitivity tests. Numbers after the site name denote whether f-no 2 has been modelled using measured () or modelled (1) roadside ozone. a) Austria >.1% of hours modelled >3% of hours modelled 3% 3% f-no2 (%) 25% 2% 15% Austria AT168A_ Austria AT21A_1 Austria AT38A_1 Austria AT24A_1 Austria AT168A_1 f-no2 (%) 25% 2% 15% Austria AT168A_ Austria AT21A_1 Austria AT38A_1 Austria AT24A_1 Austria AT168A_1 1% 1% 5% 5% % % b) Czech Republic >.1% of hours modelled >3% of hours modelled 3% 3% f-no2 (%) 25% 2% 15% Czech Republic CZ66A_1 Czech Republic CZ8A_ Czech Republic CZ13A_1 Czech Republic CZ11A_1 Czech Republic CZ65A_ Czech Republic CZ8A_1 Czech Republic CZ65A_1 f-no2 (%) 25% 2% 15% Czech Republic CZ66A_1 Czech Republic CZ8A_ Czech Republic CZ13A_1 Czech Republic CZ11A_1 Czech Republic CZ65A_ Czech Republic CZ8A_1 Czech Republic CZ65A_1 1% 1% 5% 5% % % AEA Energy & Environment 27

39 Figure 3.1 continued c) Finland >.1% of hours modelled >3% of hours modelled 3% 3% f-no2 (%) 25% 2% 15% Finland FI16A_thl met Finland FI4A_thl met Finland FI6A_thl met Finland FI16A_thl met Finland FI16A_pwa met Finland FI4A_pwa met Finland FI6A_pwa met Finland FI16A_pwa met f-no2 (%) 25% 2% 15% Finland FI16A_thl met Finland FI4A_thl met Finland FI6A_thl met Finland FI16A_thl met Finland FI16A_pwa met Finland FI4A_pwa met Finland FI6A_pwa met Finland FI16A_pwa met 1% 1% 5% 5% % % d) France >.1% of hours modelled >3% of hours modelled 3% 3% 25% France FR895A_1 25% France FR895A_1 France FR335A_1 France FR335A_1 2% France FR898A_1 2% France FR898A_1 f-no2 (%) 15% France FR91A_1 f-no2 (%) 15% France FR91A_1 1% 1% 5% 5% % % e) Germany >.1% of hours modelled >3% of hours modelled AEA Energy & Environment 28

40 3% 3% f-no2 (%) 25% 2% 15% Germany FNV_FMI Germany KAS_KAN Germany MAV_MAN Germany STS_STB Germany MAV_MAS Germany STS_STZ f-no2 (%) 25% 2% 15% Germany FNV_FMI Germany KAS_KAN Germany MAV_MAN Germany STS_STB Germany MAV_MAS Germany STS_STZ 1% 1% 5% 5% % % Figure 3.1 continued f) Greece >.1% of hours modelled >3% of hours modelled 3% 3% f-no2 (%) 25% 2% 15% Greece GR22A_ Greece GR4A_1 Greece GR32A_ Greece GR2A_ Greece GR22A_1 Greece GR32A_1 Greece GR2A_1 f-no2 (%) 25% 2% 15% Greece GR22A_ Greece GR4A_1 Greece GR32A_ Greece GR2A_ Greece GR22A_1 Greece GR32A_1 Greece GR2A_1 1% 1% 5% 5% % % g) Italy >.1% of hours modelled >3% of hours modelled AEA Energy & Environment 29

41 3% 25% 2% Italy IT477A_IT117A Italy IT771A_IT117A Italy IT116A_IT117A Italy IT75A_IT117A Italy IT477A_IT123A Italy IT771A_IT123A Italy IT116A_IT123A Italy IT75A_IT123A 3% 25% 2% Italy IT477A_IT117A Italy IT771A_IT117A Italy IT116A_IT117A Italy IT75A_IT117A Italy IT477A_IT123A Italy IT771A_IT123A Italy IT116A_IT123A Italy IT75A_IT123A f-no2 (%) 15% f-no2 (%) 15% 1% 1% 5% 5% % % h) Netherlands >.1% of hours modelled >3% of hours modelled 3% 3% f-no2 (%) 25% 2% 15% Netherlands NL248A_ Netherlands NL253A_1 Netherlands NL234A_ Netherlands NL224A_ Netherlands NL244A_1 Netherlands NL235A_ Netherlands NL248A_1 Netherlands NL234A_1 Netherlands NL224A_1 Netherlands NL235A_1 f-no2 (%) 25% 2% 15% Netherlands NL248A_ Netherlands NL253A_1 Netherlands NL234A_ Netherlands NL224A_ Netherlands NL244A_1 Netherlands NL235A_ Netherlands NL248A_1 Netherlands NL234A_1 Netherlands NL224A_1 Netherlands NL235A_1 1% 1% 5% 5% % % Figure 3.1 continued i) Spain >.1% of hours modelled >3% of hours modelled AEA Energy & Environment 3

42 3% 3% f-no2 (%) 25% 2% 15% Spain ES1453A_ Spain ES1262A_ Spain ES148A_ Spain ES1438A_ Spain ES1453A_1 Spain ES1262A_1 Spain ES148A_1 Spain ES1438A_1 f-no2 (%) 25% 2% 15% Spain ES1453A_ Spain ES1262A_ Spain ES148A_ Spain ES1438A_ Spain ES1453A_1 Spain ES1262A_1 Spain ES148A_1 Spain ES1438A_1 1% 1% 5% 5% % % j) UK >.1% of hours modelled >3% of hours modelled 3% 25% 2% UK Camden Kerbside_1 UK London A3 Roadside_1 UK London Cromwell Road 2_1 UK London Marylebone Road_1 UK Southwark Roadside_1 UK Tower Hamlets Roadside_1 UK London Marylebone Road_1 UK Camden Kerbside_1 3% UK London A3 Roadside_1 UK London Cromwell Road 2_1 UK London Marylebone Road_1 25% UK Southwark Roadside_1 UK Tower Hamlets Roadside_1 UK London Marylebone Road_1 2% f-no2 (%) 15% f-no2 (%) 15% 1% 1% 5% 5% % % AEA Energy & Environment 31

43 Table 3.2. f-no 2 for different monitoring sites and model runs. Numbers after the site name denote whether f-no 2 has been modelled using measured () or modelled (1) roadside ozone. f-no 2 Values (%) Percentage of total hours per year modelled (%) Country Site AT168A_ Austria Czech Republic Finland France AT21A_ AT38A_ AT24A_ AT168A_ CZ66A_ CZ8A_ CZ13A_ CZ11A_ CZ65A_ CZ8A_ CZ65A_ FI16A_thl met FI4A_thl met FI6A_thl met FI16A_thl met FI16A_pw a met FI4A_pw a met FI6A_pw a met FI16A_pw a met FR895A_ FR335A_ FR898A_ FR91A_ Greece GR22A_ AEA Energy & Environment 32

44 f-no 2 Values (%) Percentage of total hours per year modelled (%) Country Site GR4A_ GR32A_ Germany Italy Netherlands Spain GR2A_ GR22A_ GR32A_ GR2A_ FNV_FMI KAS_KAN MAV_MAN STS_STB MAV_MAS STS_STZ IT477A_ IT771A_ IT116A_ IT75A_ NL248A_ NL253A_ NL234A_ NL224A_ NL244A_ NL235A_ NL248A_ NL234A_ NL224A_ NL235A_ ES1453A_ ES1262A_ ES148A_ ES1438A_ ES1453A_ ES1262A_ AEA Energy & Environment 33

45 f-no 2 Values (%) Percentage of total hours per year modelled (%) Country Site ES148A_ UK ES1438A_ Camden Kerbside_ London A3 Roadside_ London Cromwell Road 2_1 London Marylebone Road_ Southwark Roadside_ Tower Hamlets Roadside_ London Marylebone Road_1 1 Model runs were done for Finland using two set of met data: HAM (Amsterdam) and PWA (Warsaw Okecie). AEA Energy & Environment 34

46 4 Emissions Analysis 4.1 Introduction This section presents the methodology used to estimate nitrogen oxide and nitrogen dioxide emissions from the road transport fleet in ten European countries. f-no 2 estimates have been calculated from these emissions estimates for 1995, 2, 25, 21, 215 and 22. The countries analysed were: Austria Czech Republic Finland France Germany Greece Italy Netherlands Spain UK Various Directives regulate nitrogen oxide emissions from the road transport sector. These are discussed in Section 4.2 below. 4.2 NO X Type Approval Limits European Directive 91/441/EC amended its parent Directive 7/22/EC and mandated emission standards for oxides of nitrogen, hydrocarbons and carbon monoxide which were sufficiently low to virtually compel vehicle manufacturers to fit three-way catalysts and electronically controlled fuel injection to all new petrol cars. This significantly reduced emissions of NO X and other pollutants. Standards for Directive 91/441/EC, frequently referred to as Euro 1, were followed by Euro 2 standards implemented by amending Directive 94/12/EC during the mid 199s. Euro 3 and Euro 4 standards were then introduced in amending Directive 98/69/EC reducing NO X emissions further. Recently NO X emission limits have been agreed, for Euro 5 and 6, which are due to be implemented in 211 and 216 respectively. NO X emissions from diesel light-duty vehicles have also been regulated by the same European directives. These emission limits, as with petrol vehicles, specify different emissions limits for passenger cars, larger passenger cars and three sizes of light-duty vans. Also, as for petrol vehicles more stringent standards have been introduced, reducing emissions of NO X. NO X emissions from heavy-duty vehicles (>3.5 tonnes gross vehicle weight) have also been regulated since the 198s, but are covered under a different family of directives (Directive 88/77/EEC and its amending directives). For these vehicles the emission standards are specified in units of g/kw hr rather than in g/km travelled. Nevertheless, the successive reduction in NO X standard with newer technologies is a feature common to these vehicles too. Amending Directive 99/96/EC specifies NO X emission limits for Euro IV and Euro V heavy duty vehicles. There is currently no agreed Euro VI for heavy-duty vehicles. The type approval limits for NO X and particulate matter (PM) by Euro standard are provided below. For information, the limits for PM are provided because marked reductions may signify the requirement by manufacturers to fit traps or other technologies to reduce PM emissions. This is likely to have a knock on effect on NO 2 emissions. This is further discussed in Section Tables present details of NO X type approval limits for different vehicle classes. AEA Energy & Environment 35

47 Table 4.1: Type approval limits for petrol passenger cars Petrol cars Diesel cars NO X standard mg/km NO X standard mg/km PM standard mg/km Date of introduction Euro /1/21 Euro /1/26 Euro /9/211 Euro /9/216 Table 4.2: Type approval limits for Heavy light-goods diesel vehicles (N1 Class 3) NO X standard mg/km PM standard mg/km Date of introduction Euro /1/21 Euro /1/26 Euro /9/211 Euro /9/216 Table 4.3: Type approval limits For heavy duty vehicles 1 NO X standard g/kwhr PM standard mg/kwhr Date of introduction Euro III /1/21 Euro IV /1/26 Euro V /1/29 Euro VI Nothing Suggested 4.3 Estimating NO X Emissions NO X emissions arising from road transport in the ten countries outlined above have been based on the European TREMOVE model (version 2.44). The model, which has been developed by Transport Mobility Leuven, is a policy assessment model, designed to study the effects of different transport and environmental policies on emissions from the transport sector. TREMOVE models both passenger and freight transport and covers the periods 1995 to 22. Due to the emission limits for Euro 5 and 6 light duty vehicles (LDVs) not being approved by the time version 2.44 of the TREMOVE model was completed, the model at present only includes LDV vehicles up to Euro 4 standard. This study however includes emission estimates from these more advanced vehicles. This data has been obtained by utilising figures that were calculated for the UK s Air Quality Strategy (AQS) work (AQS, 26). This data provided the relative change in NO X emission factors between Euro 4 and Euro 5/6 for light duty vehicles and gave information on the penetration of the fleet by these new vehicle types. Whilst the latter information is UK specific, it is thought that this data would be representative of the European fleets studied. This information is presented in Tables 4.4 and 4.5. Table 4.4. The reduction in NO X emissions expected with the introduction of Euro 5 and 6 (emissions are shown relative to a Euro 4 vehicle). Euro 5 Euro 6 Emissions relative Emissions relative Entry into service Entry into service to Euro 4 to Euro 4 Petrol cars Jan 21 66% - - Petrol LGVs Jan % - - Diesel cars Jan 21 72% Jan % Diesel LGVs Jan % Jan % Note: the table shows the expected NOx emissions relative to a Euro 4 vehicle. For example, NOx emissions from Euro 5 diesel cars are expected to be approximately 72% of a euro 4 diesel car. 1 Limit values quoted are for the ETC (European Transient Cycle) rather than ESC (European Static Cycle) +DLR (Dynamic Load Response) test cycles for heavy duty engines. AEA Energy & Environment 36

48 Table 4.5. The fraction of the fleet that originally would have conformed to Euro 4 that is now expected to conform to Euro 5 and Euro Petrol LDV Euro Petrol LDV Euro Diesel LDV Euro Diesel LDV Euro Diesel LDV Euro Note: For example, of the petrol light duty vehicles (cars + light goods vehicles) that would have conformed to euro 4 in 215, 65% of these are now expected to conform to Euro 5 standards; 35% will remain as Euro 4. For diesel LDVs in 215, of those that would have conformed to Euro 4 standards, 62% of them are now expected to conform to Euro 5 and 7% to euro 6. Using the data presented in TREMOVE and Tables 4.4 and 4.5, NO X emissions were estimated for each vehicle type by Euro standard at five yearly intervals between 1995 and 22. Using this information, total NO X emissions were then calculated for each country. This information was then combined with the nitrogen dioxide fractions presented in Section 4.4 to enable total primary nitrogen dioxide emissions to be estimated. 4.4 Estimating NO 2 Emissions Nitrogen oxides (NO X ) formed in combustion are mainly released in the form of NO. It has been assumed, especially among modellers, that 5% of the NO X released from combustion is in the form of NO 2. This has been applied to all types of combustion including power stations, domestic heaters and both petrol and diesel internal combustion engines. Recently, emission tests have been undertaken on road vehicles, showing that there are large variations in this figure depending on speed, fuel type and vehicle type. The ratio of primary nitrogen dioxide to the total oxides of nitrogen, %NO 2 to %NO X (on a volume/volume basis) leaving the vehicle s exhaust system is denoted as f-no 2 2 in this report. A number of studies on nitrogen dioxide emissions were consulted to help inform what f-no 2 figures should be used in this study. The following sections give an overview of the findings from the reports that were assessed UK Department for Transport, Primary NO 2 study (Latham et al, 21) The main objective of this research was to determine the proportion of NO X emitted as NO 2 from a range of typical vehicles with different levels of emission control technology. The data obtained from this study is presented in Figure 4.1. The key findings from the study were: Above 5 kilometres per hour (kph) f-no 2 decreases with increasing speed, whilst NO X emissions generally fall initially and then increase. Below 5 kph, changes of gear complicate the speed-related emission functions. Generally f-no 2 for Euro and Euro 1 are similar, whilst it is larger for Euro 2. The average f-no 2 is a little greater than 1% for all nine-vehicle types of the speed range 4 1 kph. 2 This is the same nomenclature as is found in AQEG (26). AEA Energy & Environment 37

49 Figure 4.1 Speed related f-no 2 for different diesel vehicle types NO2 to NOx ratios for different types of diesel vehicles 3.% 25.% NO2/NOx ratio 2.% 15.% 1.% LGV Euro LGV Euro 1 LGV Euro 2 HGV Euro HGV Euro 1 HGV Euro 2 Bus Euro Bus Euro 1 Bus Euro 2 5.%.% Average speed (kph) UK Traffic Management and Heavy Duty Vehicle Emissions Study (TRAMAQ) This study involved testing the emissions from 5 heavy goods vehicles. The conclusions of the study were: Diesel Particulate filters - 4 vehicles were fitted with traps, with two of these being also fitted with an oxidation catalyst. The effects of these technologies are inconclusive. Half of the modified vehicles had f-no 2 well below the median, and the other half well above. Exhaust Gas Recirculation (EGR) - 17 of the 5 vehicles tested were fitted with EGR systems. The report showed that EGR had no significant effect on the f-no2 ratio UK DfT Euro 3 Passenger Car Study (Collier et al, 25) This study was conducted by Ricardo on behalf of the UK s DfT. In addition to the regulated emissions tested in this research programme, NO and NO 2 emissions were collected for 12 petrol and 8 diesel cars. The project encompassed small, medium and large passenger cars. The results produced from the study showed that: For the petrol vehicles f-no 2 was less than 5%, except for the GDI vehicle, with the mean of the remaining 11 petrol vehicles being 2.3%. For the gasoline direct injection (GDI) vehicle f-no 2 was 1%. For the diesel vehicles f-no 2 ranged from 8.4% to 48.5%, with the mean value 28.3%. The study also presented f-no 2 against the average speed of the drive cycle. For the petrol vehicles, no correlation is evident. For diesel vehicles, there is a positive correlation between speed and f-no 2 (see Figure 4.2). For reasons of confidentiality we are unable to be more specific with the labelling in 4.2. AEA Energy & Environment 38

50 Figure 4.2 The NO 2 /NO X ratio as a function of speed for Euro 3 petrol cars and Euro 3 diesel cars NO 2 /N O X ratio as a function of speed for Euro 3 petrol cars 1.% 9.% NO 2 /NO X ratio (%) 8.% 7.% 6.% 5.% 4.% 3.% Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 1 Vehicle 11 Vehicle 12 2.% 1.%.% Speed (kph) NO 2 /NO X ratio as a function of speed for Euro 3 diesel cars 8.% 7.% 6.% NO 2 /NO X ratio (%) 5.% 4.% 3.% Vehicles 1 Vehicles 2 Vehicles 3 Vehicles 4 Vehicles 5 Vehicles 6 Vehicles 7 Vehicles 8 2.% 1.%.% Speed (kph) UK DfT Study on New and Emerging Technologies (Stones et al, 26) This study was undertaken by AEA and sought to quantify the effect of new technologies on emissions, especially those not currently regulated. Four emerging technologies were evaluated, a Fuel Borne Catalyst combined with a Diesel Particulate Filter (FBC+DPF), a Catalysed Diesel Particulate Filter (DPF), a Gasoline Direct Injection combined with a Lean NO X Trap (GDI+LNT) and a Diesel Particulate and NO X Reduction (DPNR) system. The vehicles were each tested over a range of drive cycles. It is important to note that these tests have been carried out on few vehicles and therefore they may not be truly representative of the national fleet. The vehicles met the Euro 4 emission standards. The study showed that for two of the four vehicles, the percentage of primary NO 2 was in the 5 to 7% range, and 2 to 5% range for the other 2 vehicles. The results from this study together with data for petrol and diesel Euro III cars from the Ricardo study are shown Figure 4.3a Figure 4.3b shows the data from Figure 4.3a plotted against the mean cycle speed, rather than simply as adjacent columns. For the two vehicles with the highest primary NO 2 ratio, the purple and bright green columns of Figure 4.3a, Figure 4.3b shows that this ratio has little dependency on speed, i.e. the high ratio is a feature of the vehicle type for all driving conditions. AEA Energy & Environment 39

51 For reasons of confidentiality we are unable to be more specific with the labelling in 4.3. Figure 4.3a The f-no 2 of some emerging diesel technologies together with data for Euro 3 petrol, diesel and GDI technology for a range of drive cycles 8% 7% Petrol - Euro 3 Diesel - Euro 3 GDI - Euro 3 New Technology 1 New technology 2 New Technology 3 ver 1 New Technology 3 ver 2 6% Percentage NO 2 of NO 2 +NO (%) 5% 4% 3% 2% 1% % ECE EUDC Cold NEDC Hot NEDC Congested Congested Urban 1 Urban 2 Artemis Urban WS Suburban WS Rural SubUrban + Rural Dbl Mway 9 Dbl Mway 113 Artemis Mway kph SS Figure 4.3b The f-no 2 from figure 1.3.4a plotted against mean drive cycle speed 8% Petrol Euro 3 Percentage NO 2 of NO2+NO (%) 7% 6% 5% 4% 3% Diesel Euro 3 GDI Euro 3 New Technology 1 New Technology 2 New Technology 3 Ver 1 New Technology 3 Ver 2 2% 1% % Average speed of drive cycle (kph) AEA Energy & Environment 4

52 4.4.5 UK transport for London (TfL) Bus Study (TfL, 26) Transport for London (TfL) have new vehicle types tested on a chassis dynamometer by Millbrook prior to their operation on London streets. The drive cycle used, the Millbrook London Transport bus cycle, was developed from a real world bus cycle within London. The data for Euro III buses (fitted with a particulate trap supplied by Eminox) gave f-no 2 around 4% ± 1%. For two Euro IV buses tested, both using selective catalytic reduction (SCR) to control emissions, f-no 2 was around 6 ± 2%, markedly lower than for the vehicles fitted with traps International Studies In addition to the UK studies discussed in the earlier sections of this chapter, a number of relevant international studies have also been undertaken. However, the results of some of these have not been published in the public domain. Consequently, the results are not available for analysis. General trends and conclusions have been obtained by talking with the organisations involved. Study at LAT The Laboratory of Applied Thermodynamics at the University of Thessaloniki studied the f-no 2 for four passenger cars. This gave f-no 2 in the range 25 8% for the diesel vehicles dependent on drive cycle. TNO EMPA study This study comprised of testing the emissions of 69 passenger cars. The main findings with regard to f-no 2 as a function of technology were: For petrol vehicles f-no 2 was found to be up to approximately 9% but 17% for GDI vehicles, For diesel passenger cars f-no 2 had a mean value of around 18% for pre-euro 1, Euro 1 and Euro 2 vehicles For diesel passenger cars f-no 2 rises to around 5% for Euro 3 and some Euro 4 cars, but with some technologies having an f-no 2 of around 7%. These data are in good agreement with the UK DfT study on emerging technologies. Dutch Ministry of Environment bus study The key findings of this study were that the f-no 2 for Euro II buses without DPFs ranged from 6% to 8%, whereas for Euro II vehicles retrofitted with CRT it ranged form 9% to 22%. Euro III vehicles equipped with OEM DPF systems averaged 12%. These data are significantly lower than those recorded from the UK s TfL measurement programme Recommended f-no 2 From the data reviewed, it is clear that the assumption of a 5% f-no 2 rate is a systematic underestimate for diesel vehicles. Furthermore, the data strongly indicates that no single value of f- NO 2 is appropriate for all vehicle types. Rather, it is dependent on: Vehicle type (passenger car, heavy goods vehicle, bus etc) Emissions standard Any exhaust after treatment fitted The average speed of the drive cycle. On the basis of these data, the baseline emission factors used in this study are listed in Table 4.6. For diesel vehicles the general trends are a reduction in NO X emissions with accompanying major changes in f-no 2 caused by the addition of exhaust after-treatment systems. For the heavy-duty vehicles this is seen between Euro II and Euro III where the fitting of diesel particulate traps causes an increase in f-no 2. However, the fraction of vehicles affected varies from close to 1% for buses and coaches (captive fleets which operate principally in urban areas) to around 1% for trucks. Consequently, the size of the increase in f-no 2 varies systematically between trucks, buses and coaches, despite the fact that they use common types of heavy-duty engines. AEA Energy & Environment 41

53 A further tenet upon which these baseline data are founded is that selective catalytic reduction (SCR) will be the dominant exhaust after-treatment system fitted to Euro IV and later heavy-duty vehicles and that f-no2 for SCR is 1% (although it is acknowledged that this is based on only a little experimental data). Table 4.6. Measured NOx emission factors and recommended f-no 2 Petrol passenger cars Diesel passenger cars Diesel heavy lightduty vans Rigid trucks Articulated trucks NO X (g/km) f-no 2 (%) Speed related f-no 2 (low, medium & high) Euro 2 and earlier.25 4 Euro Euro Euro Euro Euro 2 and earlier.7 11 Euro %, 3%, 4% Euro Euro Euro Euro 2 and earlier Euro Euro Euro Euro Euro II and earlier Euro III %, 14%, 8% Euro IV 3. 1 Euro V Euro VI Euro II and earlier Euro III %, 14%, 8% Euro IV Euro V 4. 1 Euro VI 4. 1 Euro II and earlier Euro III untrapped Euro III trapped 1 35 Buses and coaches Euro IV 7 1 Euro V 4 1 Euro VI 4 1 Note: - Speed related f-no 2 are only provided for those vehicles that have shown clear f-no 2 speed dependency. For other vehicle types, the f-no 2 has been assumed to be the same at all speeds. - NO2 emission factors are only shown in the above table for information only. NO X emissions have been taken from the TREMOVE model for this study. The f-no 2 data provided in Table 4.6 has been combined with emissions data from TREMOVE and information from the UK s Air Quality Strategy (AQS) to estimate NO 2 emissions for the 1 Member States included in this study. The estimated historical and projected NO X and NO 2 emissions and the f-no 2 are presented for the base case in Tables 4.7 to AEA Energy & Environment 42

54 Table 4.7. Estimated total urban NO X emissions arising from road transport activity (ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Table 4.8. Estimated total nationwide NO X emissions arising from road transport activity (Ktons) Country UK 1, France Germany 1,172 1, Spain Greece Italy Czech Republic Austria Netherlands Finland Table 4.9. Estimated average f-no 2 for urban areas (%) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Table 4.1. Estimated average f-no 2 nationwide (%) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland AEA Energy & Environment 43

55 Table Estimated total urban NO 2 emissions arising from road transport activity (ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Table Estimated total nationwide NO 2 emissions arising from road transport activity (ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland The data shows that NO X emissions from road transport are expected to decline substantially by 22. However, f-no 2 is predicted to rise considerably in the future, primarily due to the fitting of exhaust after-treatments, which lead to a higher proportion of the NO X being emitted as NO 2. The exception to this is in Greece where only a small rise in the overall f-no 2 is predicted. This is due to no private diesel cars being allowed in Athens or Thessaloniki, which results in a large proportion of NO X emissions being from petrol vehicles, which have the lowest f-no 2. f-no 2 is generally predicted to increase steeply from 25 to 215. Thus urban NO 2 emissions are predicted to increase from 2 to 21 in contrast to the decline in NO X emissions. Urban NO 2 emissions are then predicted to flatten off to 215 and then decline to roughly equivalent to 25 values in 22 for the baseline. By 22 the decease in NO X emissions is sufficient to offset the increase in f-no Scenarios In addition to baseline emissions being estimated, four further scenarios were developed. They were as follows: Scenario 1: Heavy duty vehicle pessimistic scenario The key difference relative to the baseline scenario is that SCR does not cause a reduction in f- NO 2. Consequently the f-no 2 ratio for all three heavy-duty vehicle types remains at the Euro III level rather than decreases. It is also assumed that no Euro VI standard is agreed for heavy-duty vehicles. Scenario 2: Heavy duty vehicle optimistic scenario In this scenario the two key differences, relative to the baseline scenario, are: AEA Energy & Environment 44

56 a) The effect of SCR on f-no 2 is somewhat better than the conservative interpretation of the few existing measurements. Consequently, the f-no 2 ratio for all heavy-duty vehicles fitted with SCR (Euro IV and later) is less than that presumed in the baseline case, b) There is a bold reduction required for Euro VI (along the lines of the US HDV regulations) introduced in 1/1/213 for new models, 1/1/214 for all HDVs. This reduces the NO X emissions greatly; with f-no 2 remaining at the low 6% level for SCR equipped vehicles. Scenario 3: Changes to NO X emissions for light duty vehicles NO X emission factors and f-no 2 ratios for all heavy-duty vehicles are unchanged from the baseline scenario. However, the changes in NO X emissions for light-duty diesel vehicles (as specified in the Euro 5 and Euro 6 standards and reflected in the NO X emission factors for the baseline case) are achieved by the addition of SCR systems from Euro 5 for the heavier diesel light-duty vans, and from Euro 6 for all diesel light-duty vehicles. This causes f-no 2 to fall to 1% for these future emission standards. (There are possibilities that diesel cars will not use SCR but instead will fit NO X traps. The f-no 2 from such vehicles is not known, but is likely to be higher than SCR fitted vehicles). Scenario 4: Heavy duty and light duty optimistic scenario As for scenario 2 for heavy duty vehicles and lower f-no 2 (6%) for light duty vehicles than in Scenario 3. As with Scenario 2 there is a bold reduction required for Euro VI NO X (along the lines of the US HDV regulations) for heavy duty vehicles. The representative NO X emissions and estimated f-no 2 under the different scenarios are presented in Table Those cells highlighted in gold indicate where there are differences from the baseline. AEA Energy & Environment 45

57 Table 4.13 Representative NO X emissions and estimated f-no 2 under the different scenarios Petrol passenger cars Diesel passenger cars Diesel heavy lightduty vans Rigid trucks Articulated trucks Buses and coaches Baseline Scenario 1 Scenario 2 Scenario 3 Scenario 4 NO X (g/km) f-no 2 NO X (g/km) f-no 2 NO X (g/km) f-no 2 NO X (g/km) f-no 2 NO X (g/km) f-no 2 Euro 2 and earlier.25 4%.25 4%.25 4%.25 4%.25 4% Euro 3.5 3%.5 3%.5 3%.5 3%.5 3% Euro 4.3 3%.3 3%.3 3%.3 3%.3 3% Euro 5.3 3%.3 3%.3 3%.3 3%.3 3% Euro 6.3 3%.3 3%.3 3%.3 3%.3 3% Euro 2 and earlier.7 11%.7 11%.7 11%.7 11%.7 11% Euro 3.6 3%.6 3%.6 3%.6 3%.6 3% Euro %.4 55%.4 55%.4 55%.4 55% Euro %.3 55%.3 55%.3 55%.3 55% Euro %.1 55%.1 55%.1 1%.1 6% Euro 2 and earlier % % % % % Euro % 1. 3% 1. 3% 1. 3% 1. 3% Euro %.7 55%.7 55%.7 55%.7 55% Euro %.5 55%.5 55%.5 1%.5 6% Euro %.2 55%.2 55%.2 1%.2 6% Euro II and earlier 7. 11% 7. 11% 7. 11% 7. 11% 7. 11% Euro III 4. 14% 4. 14% 4. 14% 4. 14% 4. 14% Euro IV 3. 1% 3. 14% 3. 6% 3. 1% 3. 6% Euro V 1.7 1% % 1.7 6% 1.7 1% 1.7 6% Euro VI 1.7 1% %.3 6% 1.7 1%.3 6% Euro II and earlier % % % % % Euro III 9. 14% 9. 14% 9. 14% 9. 14% 9. 14% Euro IV 6.5 1% % 6.5 6% 6.5 1% 6.5 6% Euro V 4. 1% 4. 14% 4. 6% 4. 1% 4. 6% Euro VI 4. 1% 4. 14%.7 6% 4. 1%.7 6% Euro II and earlier 15 11% 15 11% 15 11% 15 11% 15 11% Euro III no trap 11 14% 11 14% 11 14% 11 14% 11 14% Euro III trapped 1 35% 1 35% 1 35% 1 35% 1 35% Euro IV 7 1% 7 35% 7 6% 7 1% 7 6% Euro V 4 1% 4 35% 4 6% 4 1% 4 6% Euro VI 4 1% 4 35%.7 6% 4 1%.7 6% Note: NO X emission factors are shown for information only. Data from TREMOVE has been used in this study. AEA Energy & Environment 46

58 4.5.3 Scenario Results Tables 4.14 to 4.17 present the estimated NO X and NO 2 emissions nationwide and in urban areas and the corresponding f-no 2 in each of the ten Member States analysed for each of the scenarios. Figure 4.4 provides a summary of the f-no 2 scenario results for urban areas. The f-no 2 in 21, 215 and 22 increases under scenario 1 compared to the baseline. This is as a result of the predicted f-no 2 increasing for Euro IV, V and VI buses from 1% to 35%. The estimated NO X emissions remain the same as in the baseline. Urban and nationwide NO 2 emissions are therefore predicted to be slightly higher for scenario 1 than for the baseline. For example, by 22, across all 1 Member States considered, scenario 1 is predicted to result in urban NO 2 road transport emissions of 6kt compared with 56kt for the baseline Under Scenario 2, the f-no 2 tends to be higher than the baseline but lower than scenario 1. This is as a result of the introduction of strict NO X standards for heavy-duty vehicles (Euro VI). This results in substantially reduced NO X emissions compared to the baseline both in urban areas and nationwide and this leads to an increase in the average f-no 2. Urban NO 2 emissions for this scenario are predicted to slightly less than the baseline, but closely following the same trend as the baseline. Nationwide emissions are predicted to show more substantial deviation from the baseline and to be more similar in magnitude to scenario 3 than the baseline. In scenario 3, Euro 5 diesel vans and Euro 6 diesel passenger cars have SCR fitted. The estimated NO X emissions both nationwide and in urban areas are the same as in the baseline but the f-no 2 decreases due to manufacturers using SCR to meet the stringent NO X limits which results in lower f- NO 2 than other technologies. The lower f-no 2 levels result in a significant reduction in absolute emissions of NO 2 relative to the baseline for both urban and nationwide road traffic sources. For urban roads this scenario is predicted to result in emissions in 22 lower than the urban roads emission in 25 for most of the Member States considered. Under Scenario 4, f-no 2 values are substantially lower than in the baseline case or scenario 1 or 2 but only slightly lower than that achieved by scenario 3. This is due to the lower f-no 2 assumed for light duty vehicles. This scenario is therefore predicted to be the most effective at reducing emissions of NO 2 relative to the baseline for both urban and nationwide roads. AEA Energy & Environment 47

59 Table Scenario 1: HDV pessimistic scenario Total urban NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Total nationwide NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Estimated average f-no 2 for urban areas Country UK 5.4% 5.9% 1.2% 21.5% 33.3% 38.2% France 5.8% 6.9% 13.9% 26.4% 36.9% 42.2% Germany 5.5% 6.% 1.1% 2.1% 29.1% 33.3% Spain 5.8% 6.5% 1.5% 19.% 28.% 33.8% Greece 6.9% 6.9% 7.3% 8.1% 9.2% 1.7% Italy 5.9% 6.4% 8.8% 15.1% 22.3% 27.5% Czech Republic 6.1% 6.3% 9.8% 15.8% 21.5% 21.3% Austria 6.5% 7.3% 11.1% 21.2% 31.4% 36.5% Netherlands 5.1% 5.6% 8.6% 17.6% 28.% 28.9% Finland 4.8% 5.9% 7.1% 1.7% 19.7% 25.6% Estimated average f-no 2 nationwide Country UK 6.4% 7.3% 11.9% 2.4% 28.7% 3.1% France 7.5% 8.7% 13.7% 2.4% 26.% 26.8% Germany 7.4% 8.2% 11.3% 16.% 2.1% 21.2% Spain 8.6% 9.2% 11.7% 15.6% 19.5% 2.5% Greece 7.8% 8.% 8.5% 9.% 1.% 11.2% Italy 7.% 7.6% 1.3% 15.6% 21.2% 23.5% Czech Republic 8.1% 8.5% 11.9% 15.2% 17.4% 17.2% Austria 8.% 8.8% 12.6% 18.3% 23.3% 23.9% Netherlands 7.1% 7.8% 11.3% 16.8% 21.9% 21.4% Finland 7.2% 8.2% 1.% 12.9% 17.8% 2.4% Total urban NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland AEA Energy & Environment 48

60 Total nationwide NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Table Scenario 2: HDV optimistic scenario Total urban NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Total nationwide NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Estimated average f-no 2 for urban areas Country UK 5.4% 5.9% 1.2% 2.7% 32.3% 39.2% France 5.8% 6.9% 13.9% 25.8% 36.2% 41.3% Germany 5.5% 6.% 1.1% 19.2% 27.8% 31.8% Spain 5.8% 6.5% 1.5% 18.7% 27.6% 33.6% Greece 6.9% 6.9% 7.3% 6.6% 6.9% 7.% Italy 5.9% 6.4% 8.8% 14.4% 21.% 25.5% Czech Republic 6.1% 6.3% 9.8% 14.8% 2.2% 2.9% Austria 6.5% 7.3% 11.1% 19.7% 29.2% 33.3% Netherlands 5.1% 5.6% 8.6% 17.2% 27.1% 28.% Finland 4.8% 5.9% 7.1% 1.1% 18.4% 23.6% AEA Energy & Environment 49

61 Estimated average f-no 2 nationwide Country UK 6.4% 7.3% 11.9% 18.7% 26.7% 33.1% France 7.5% 8.7% 13.7% 19.3% 25.% 28.7% Germany 7.4% 8.2% 11.3% 15.1% 18.7% 2.5% Spain 8.6% 9.2% 11.7% 14.4% 17.8% 2.5% Greece 7.8% 8.% 8.5% 8.2% 8.2% 8.3% Italy 7.% 7.6% 1.3% 14.7% 19.8% 22.8% Czech Republic 8.1% 8.5% 11.9% 13.8% 15.5% 15.9% Austria 8.% 8.8% 12.6% 17.% 21.5% 23.5% Netherlands 7.1% 7.8% 11.3% 15.3% 19.7% 21.5% Finland 7.2% 8.2% 1.% 11.9% 15.9% 19.6% Total urban NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Total nationwide NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Table Scenario 3: changes to NO X emissions for LDV Total urban NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland AEA Energy & Environment 5

62 Total nationwide NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Estimated average f-no 2 for urban areas Country UK 5.4% 5.9% 1.2% 2.8% 27.1% 2.2% France 5.8% 6.9% 13.9% 25.8% 34.2% 28.9% Germany 5.5% 6.% 1.1% 19.3% 26.9% 23.4% Spain 5.8% 6.5% 1.5% 18.7% 26.1% 24.5% Greece 6.9% 6.9% 7.3% 6.9% 7.3% 7.3% Italy 5.9% 6.4% 8.8% 14.5% 2.5% 19.3% Czech Republic 6.1% 6.3% 9.8% 15.% 18.4% 14.7% Austria 6.5% 7.3% 11.1% 19.9% 28.6% 24.3% Netherlands 5.1% 5.6% 8.6% 17.3% 26.6% 2.3% Finland 4.8% 5.9% 7.1% 1.1% 16.2% 15.9% Estimated average f-no 2 nationwide Country UK 6.4% 7.3% 11.9% 19.1% 22.3% 16.9% France 7.5% 8.7% 13.7% 19.7% 22.9% 18.5% Germany 7.4% 8.2% 11.3% 15.4% 18.1% 15.6% Spain 8.6% 9.2% 11.7% 14.9% 17.% 15.% Greece 7.8% 8.% 8.5% 8.5% 8.8% 8.9% Italy 7.% 7.6% 1.3% 15.% 19.2% 17.% Czech Republic 8.1% 8.5% 11.9% 14.4% 15.% 12.8% Austria 8.% 8.8% 12.6% 17.5% 21.1% 17.2% Netherlands 7.1% 7.8% 11.3% 15.9% 19.7% 15.1% Finland 7.2% 8.2% 1.% 12.2% 14.3% 13.2% Total urban NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Total nationwide NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland AEA Energy & Environment 51

63 Table Scenario 4: HDV and LDV optimistic Total urban NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Total nationwide NO X emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland Estimated average f-no 2 for urban areas Country UK 5% 6% 1% 2% 27% 2% France 6% 7% 14% 26% 34% 28% Germany 6% 6% 1% 19% 27% 23% Spain 6% 6% 1% 19% 26% 24% Greece 7% 7% 7% 7% 7% 7% Italy 6% 6% 9% 14% 2% 19% Czech Republic 6% 6% 1% 15% 18% 15% Austria 7% 7% 11% 2% 29% 24% Netherlands 5% 6% 9% 17% 27% 2% Finland 5% 6% 7% 1% 16% 15% Estimated average f-no 2 nationwide Country UK 6% 7% 12% 19% 22% 18% France 8% 9% 14% 19% 23% 2% Germany 7% 8% 11% 15% 18% 16% Spain 9% 9% 12% 14% 17% 16% Greece 8% 8% 8% 8% 8% 8% Italy 7% 8% 1% 15% 19% 17% Czech Republic 8% 8% 12% 14% 14% 13% Austria 8% 9% 13% 17% 21% 18% Netherlands 7% 8% 11% 15% 19% 16% Finland 7% 8% 1% 12% 14% 13% Total urban NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland AEA Energy & Environment 52

64 Total nationwide NO 2 emissions (Ktons) Country UK France Germany Spain Greece Italy Czech Republic Austria Netherlands Finland AEA Energy & Environment 53

65 Figure 4.4. Graphs showing estimated f-no 2 for the baseline and the scenarios a) Austria b) Czech Republic The estimated f-no2 in urban areas in Austria The estimated f-no2 in urban areas in the Czech Republic f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 c) Finland d) France The estimated f-no2 in urban areas in Finland The estimated f-no2 in urban areas in France f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 e) Greece f) Germany The estimated f-no2 in urban areas in Greece The estimated f-no2 in urban areas in Germany f-no2 (%) 12% 1% 8% 6% 4% 2% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 g) Italy h) Netherlands AEA Energy & Environment 54

66 The estimated f-no2 in urban areas in Italy The estimated f-no2 in urban areas in the Netherlands f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 Figure 4.4. continued i) Spain j) UK The estimated f-no2 in urban areas in Spain The estimated f-no2 in urban areas in the UK f-no2 (%) 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (%) 45% 4% 35% 3% 25% 2% 15% 1% 5% % Basecase Scenario 1 Scenario 2 Scenario 3 Scenario Summary Figure 4.5 summarises the emission projections for urban road traffic for the different scenarios in terms of total NO X and NO 2 emissions and average f-no 2 across the ten Member States considered. NO X emissions decline steeply to 22 and are lowest for scenarios 2 and 4 in 22. f-no 2 is predicted to increase steeply from 25 to 215 and then starting to decline for scenarios 3 and 4. Thus NO 2 emissions are predicted to increase from 2 to 21 in contrast to the decline in NO X emissions. NO 2 emissions are then predicted to flatten off to 215 and then decline to roughly equivalent to 25 values in 22 for the baseline and scenarios 1 and 2. By 22 the decease in NO X emissions is sufficient to offset the increase in f-no 2. NO 2 emissions are projected to decrease more steeply to 22 for scenarios 3 and 4 to values below 2 emissions as a result of a decrease in f-no 2 combined with NO X emission reductions. AEA Energy & Environment 55

67 Figure 4.5. Graphs summarising urban road traffic NO X and NO 2 emissions summed across the ten member states considered along with average urban f-no 2 for the different emission projection scenarios a) NO x emissions (ktonnes per year) b) NO 2 emissions (ktonnes per year) Baseline Scenario 1 Scenario 2 Scenario 3 Scenario Baseline Scenario 1 Scenario 2 Scenario 3 Scenario ktonnes per year ktonnes per year year year c) f-no 2 (percent) 4% 35% 3% Baseline Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (percent) 25% 2% 15% 1% 5% % year AEA Energy & Environment 56

68 5 Comparison of f-no 2 from the Netcen Primary NO 2 model and Emissions Analysis for Recent s 5.1 Introduction In the previous sections, f-no 2 for recent years has been estimated using two methods: The Netcen Primary NO 2 model (section 3) Analysis of emissions data (section 4) This section presents a comparison of the results for these two alternative approaches to calculating f- NO 2 and attempts to draw conclusions regarding how f-no 2 has changed over the period of interest for each of the case study areas. Comparing the Netcen Primary NO 2 model results with the emissions analysis results is not quite a case of comparing like with like. This is because the Netcen primary NO 2 model is designed to calculate f-no 2 for a specific point on a given road next to a roadside monitoring site, while the emissions analysis is a national scale average f-no 2. Both scales are useful to consider in this analysis. The local nature of the Netcen Primary NO 2 model gives detailed information on trends for roads where there are known to be problems with high ambient NO 2 concentrations. These are likely to be the roads where the 21 NO 2 limit values will be hardest to meet, so it is useful to get a good understanding of how f-no 2 is varying at these specific locations. By contrast, the emissions analysis gives a wider perspective on recent trends in f-no 2. This is particularly useful in extrapolating the results and conclusions drawn here across the wider EU area. Two entirely separate approaches using different data sources have been adopted to calculate f-no 2 so a comparison of the results also provides some verification of these approaches. This is because if the results and trends are broadly similar, despite the different methods used, this strengthens the confidence we can put in the conclusions drawn here. Where the results for the two approaches are significantly different, there may be local factors at work as discussed below. 5.2 Results Comparison Figure 5.1 presents summary graphs showing our best estimate of f-no 2 for each case study area using the Netcen Primary NO 2 model (black and red lines) compared with the national scale emissions results for the country in which the case study area is located (blue lines). The emissions results are divided into Average Nationwide, which includes vehicles on all types of roads, and Urban, which includes emissions from vehicles on urban roads only. The urban category is more directly comparable with the modelling results from the Netcen Primary NO 2 model than the nationwide results because the roadside monitoring sites analysed using the Netcen Primary NO 2 model are all located in urban areas. Results from the Netcen Primary NO 2 model are only shown where >3% of total hours for any given year have been modelled. The justification for choosing this cut off threshold is given in section Austria In Austria (Figure 5.1a), f-no 2 modelled using the Netcen primary NO 2 model at AT38A_1 (Salzburg Rudolfsplatz) closely corresponds to the f-no 2 calculated for urban roads. AT168A_1 (Salzburg Mirabellplatz) has a modelled f-no 2 of up to 4% less than the urban line. The average nationwide f-no 2 is higher than for all other results shown. Both methods for calculating f-no 2 suggest an upward trend in f-no 2 between 2 and 24/25. AEA Energy & Environment 57

69 5.2.2 Czech Republic For the Czech Republic (Figure 5.1b), there is good correlation between the urban emissions and the Netcen primary NO 2 model results. This suggests good agreement between the two modelling approaches. The scatter of the different roadside sites around the urban line shows that f-no 2 varies locally and the characteristics of different roads within Prague mean that a whole range of f-no 2 values will be found within this area Finland Very few of the monitoring sites in the Finland case study (Figure 5.1c) had a sufficient number of hours modelled to be presented on the graph. This means that the emissions analysis here is particularly useful in providing information on trends in f-no 2. The higher of the two emissions calculations of f-no 2, the average for nationwide roads, is relatively low compared with most of the other case studies considered. This, combined with the relatively low measured NO X and NO 2 concentrations at these sites, suggests that there is not currently a significant issue with primary NO 2 within Finland France In France (Figure 5.1d), the f-no 2 calculated using emissions on urban roads closely reflects the f- NO 2 calculated at various roadside monitoring sites in Paris using the Netcen Primary NO 2 model between 1995 and 2. However, from 2 to 25, the upward trend of the emissions urban line is slightly steeper than the lines calculated using the Necen Primary NO 2 model Greece In Greece (Figure 5.1e), the emissions results suggest a very slight upward trend in f-no 2 over the period considered. This trend is less steep than in the other case study Member States and results from the low number of diesel light duty vehicles in Greece. The results of modelling using the Netcen Primary NO 2 model results suggest that in general the emissions results are in the right ballpark, without really giving any clear indication of trends themselves Germany The emissions of f-no 2 results for Germany (Figure 5.1f) agree well with the Netcen Primary NO 2 model results for a selection of roadside sites in Baden Wűrttemberg between 1995 and However after this date, the modelled f-no 2 at the sites in Baden Wűrttemberg increased more rapidly than the emissions results suggest for the whole of Germany. This is thought to be a result of buses with particulate filters being introduced to the vehicle fleet in Baden Wűrttemberg. The lower f-no 2 at MAV (Manheim Strassenstation) is a reflection of changes in traffic flow due to local construction work in 24 and 25 (Werner Scholtz pers. comm. 27), and the fact that it closely matches the nationwide emissions calculations for f-no 2 is coincidental Italy In Italy (Figure 5.1g), the f-no 2 calculated using the Netcen Primary NO 2 model at a selection of roadside sites in Milan shows a very wide scatter, with no clear overall trend across the sites considered. Both the average nationwide roads and the urban roads emissions calculations of f-no 2 lie within the range of results modelled using the Netcen Primary NO 2 model. The emissions results suggest and upward trend in f-no 2. AEA Energy & Environment 58

70 5.2.8 Netherlands For the Netherlands case study (Figure 5.1h), the majority of modelled f-no 2 values for individual roadside sites using the Netcen Primary NO 2 model have an f-no 2 somewhere within the range of f- NO 2 values calculated for the two emissions categories. This means that the urban roads emissions based calculations of f-no 2 under predict f-no 2 compared with at the individual monitoring sites Spain As with Milan, the f-no 2 calculated using the Netcen Primary NO 2 model at a selection of roadside sites in Barcelona (Figure 5.1i) shows a wide scatter with little evidence of a clear overall trend in f- NO 2. The emissions results for Spain (both urban roads only and average nationwide) are at the low end of the range of f-no 2 calculated using the Netcen primary NO 2 model despite showing a slight increase with time. Due to difficulties associated with modelling f-no 2 in the Barcelona case study where the background monitoring site used is a long way from the city centre and the ozone climate is such that a lot of ozone chemistry may be occurring between this background site and the roadside sites, the emissions calculations of f-no 2 are probably more reliable in this particular situation UK In the UK case study (Figure 5.1j) both approaches adopted here show an upward trend in f-no 2. As with many of the other case studies, the results from the Netcen Primary NO 2 model show significant scatter due to the differing characteristics of the roads on which the selected monitoring sites are located. By 22, the Netcen Primary NO 2 model f-no 2 calculations were all above the urban emissions calculation of f-no 2. This difference is likely to be due to geographical differences in f-no 2 across the UK. AQEG (26) have concluded, based on emissions calculations, that f-no 2 in London is higher than across the UK as a whole and that f-no 2 is higher in central London that outer London due to buses fitted with particulate traps and the larger proportion of light duty vehicles that are taxis and therefore have diesel engines. It is unclear why the f-no 2 at London Cromwell Road 2 (GB695A) is so high. At London Marylebone Road (GB682A), the step change increase in f-no 2 in 23 is thought to be the result of the introduction of the congestion charge altering traffic flows and the introduction of more buses with fitted particulate traps. AEA Energy & Environment 59

71 Figure 5.1. Graphs comparing f-no 2 calculated using the Netcen Primary NO 2 model with >3% of hours per year modelled (black and red lines) and f-no 2 from emissions analysis (blue lines) for each case study area/country considered a) Austria b) Czech Republic f-no2 (%) 3% 25% 2% 15% Austria AT168A_ Austria AT21A_1 Austria AT38A_1 Austria AT24A_1 Austria AT168A_1 Austria Urban Austria Average Nationwide f-no2 (%) 3% 25% 2% 15% Czech Republic CZ66A_1 Czech Republic CZ8A_ Czech Republic CZ13A_1 Czech Republic CZ11A_1 Czech Republic CZ65A_ Czech Republic CZ8A_1 Czech Republic CZ65A_1 Czech Republic Urban Czech Republic Average Nationwide 1% 1% 5% 5% % % c) Finland d) France f-no2 (%) 3% 25% 2% 15% Finland FI16A_thl met Finland FI4A_thl met Finland FI6A_thl met Finland FI16A_thl met Finland FI16A_pwa met Finland FI4A_pwa met Finland FI6A_pwa met Finland FI16A_pwa met Finland Urban Finland Average Nationwide f-no2 (%) 3% 25% 2% 15% France FR895A_1 France FR335A_1 France FR898A_1 France FR91A_1 France Urban France Average Nationwide 1% 1% 5% 5% % % e) Greece f) Germany AEA Energy & Environment 6

72 f-no2 (%) 3% 25% 2% 15% Greece GR22A_ Greece GR4A_1 Greece GR32A_ Greece GR2A_ Greece GR22A_1 Greece GR32A_1 Greece GR2A_1 Greece Urban Greece Average Nationwide f-no2 (%) 3% 25% 2% 15% Germany FNV_FMI Germany KAS_KAN Germany MAV_MAN Germany STS_STB Germany MAV_MAS Germany STS_STZ Germany Urban Germany Average Nationwide 1% 1% 5% 5% % % Figure 5.1. continued g) Italy h) Netherlands f-no2 (%) 3% 25% 2% 15% Italy IT477A_IT117A Italy IT771A_IT117A Italy IT116A_IT117A Italy IT75A_IT117A Italy IT477A_IT123A Italy IT771A_IT123A Italy IT116A_IT123A Italy IT75A_IT123A Italy Urban Italy Average Nationwide f-no2 (%) 3% 25% 2% 15% Netherlands NL248A_ Netherlands NL253A_1 Netherlands NL234A_ Netherlands NL224A_ Netherlands NL244A_1 Netherlands NL235A_ Netherlands NL248A_1 Netherlands NL234A_1 Netherlands NL224A_1 Netherlands NL235A_1 Netherlands Urban Netherlands Average Nationwide 1% 1% 5% 5% % % i) Spain j) UK AEA Energy & Environment 61

73 f-no2 (%) 3% 25% 2% 15% Spain ES1453A_ Spain ES1262A_ Spain ES148A_ Spain ES1438A_ Spain ES1453A_1 Spain ES1262A_1 Spain ES148A_1 Spain ES1438A_1 Spain Urban Spain Average Nationwide f-no2 (%) 3% 25% 2% 15% UK Camden Kerbside_1 UK London A3 Roadside_1 UK London Cromwell Road 2_1 UK London Marylebone Road_1 UK Southwark Roadside_1 UK Tower Hamlets Roadside_1 UK London Marylebone Road_1 UK Urban UK Average Nationwide 1% 1% 5% 5% % % Conclusions from Analysis of Recent Trends in f-no 2 The results of our analysis of recent trends in f-no 2 presented above show that in most, if not all, countries considered, f-no 2 has increased over the past five to ten years. The emissions results for all countries and the results of the Netcen Primary NO 2 modelling in Paris, Baden Wűrttemberg and London show that the rate of increase in f-no 2 has generally increased since 2 compared with the rate of increase in f-no 2 between 1995 and 2. A comparison of the local scale Netcen Primary NO 2 model results with the national scale emissions results shows the following. Local factors (e.g. characteristics of individual roads) can have a significant impact on the f-no 2 for a specific road. Evidence for this can be seen in the scatter of f-no 2 modelling results for different roads within a single case study area. Additionally, regional factors can impact on f-no 2. For example both in Baden Wűrttemberg and in London, differences in the regional and national average vehicle fleets and number of buses fitted with particulate filters have caused f- NO 2 to be above the national average as calculated using national scale emissions analysis. In terms of model verification, the emissions results and the Netcen Primary NO 2 model results broadly agree which therefore suggests that we can have confidence in the results they are producing. However, limitations of the Netcen Primary NO 2 model in southern members states - Italy, Greece and Spain are highlighted by this comparison as the model does not accommodate the different photochemistry associated with these in southern ozone climates. The emissions analysis in these cases is particularly useful in providing an indication of how f-no 2 has changed in the past 5-1 years. The results of the comparison of f-no 2 estimates are further summarised in Table 5.1. This table shows both the emissions inventory based estimates of f-no 2 for urban areas for the baseline scenario and the range of f-no 2 derived from monitoring data. An estimate of f-no 2 for each site has been included where data capture is greater than 3%. The reason for showing a range rather than an average value is that we have only considered 4-6 roadside monitoring sites within each case study area. Therefore, while these sites do represent a range of local conditions, they are not necessarily representative of a wide area. Table 5.1. Summary of f-no 2 for each member state derived from both ambient measurements and emission inventory calculations (baseline scenario) AEA Energy & Environment 62

74 Country Method Austria Czech Republic Finland France Germany Greece Italy Netherland s Spain UK Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions Measurements Emissions AEA Energy & Environment 63

75 6 Future Ambient NO 2 Concentrations: Baseline projections 6.1 Introduction Module 3: the projection module of the Netcen Primary NO 2 model is used here to estimate future concentrations of ambient NO 2 at the roadside sites selected for this analysis. Box 1.1 (see section 1.3) gives a brief description of this module. Further details are given in Abbott (26). This section presents details of this modelling for baseline conditions. This includes: A description of assumptions made in projecting the model forwards for future years under baseline conditions Verification of the model using base year concentration measurements Calculations of model projections of hourly NO 2 for 21, 215 and 22 for baseline conditions. These hourly data are then used to calculate projected annual mean NO 2 concentrations for these years. A sensitivity analysis of the baseline projections to changes in various input parameters including the model sensitivity to total regional oxidant and f-no 2 is presented in section 7. Section 8 gives details of modelling of ambient NO 2 concentrations for 21, 215 and 22 for a range of scenarios. These scenarios have been selected to assess the likely impact on future ambient NO 2 concentrations of possible variations in future the Euro standards yet to be agreed and the impact of the technologies adopted by vehicle manufacturers to meet these standards. 6.2 Modelling Assumptions The projection module of the Netcen Primary NO 2 model requires several input parameters to run. Details of the data needed and how this data has been generated, including key assumptions that have been made, are set out below. Where emissions data have been used in calculating input parameters below, we have assumed that 25 data is representative of has been used because neither TREMOVE 2.44 nor GAINS (26) have 24 data available f-no 2 Projected f-no 2 for the year being modelled at any given site is needed. For this analysis, we have adopted an emissions inventory approach to calculating f-no 2 (see section 4 for further details of this approach). This approach results in one f-no 2 value for urban roads in each case study country for each of 25 (24 is assumed to also have this f-no 2 ), 21, 215 and 22. This is the f-no 2 that we have used at each roadside site included in the case study for that country for these years. The main advantage of using this approach to generate a f-no 2 term (and the reason for adopting this approach here) is that projected changes to the make up of the vehicle fleet and to vehicle technologies, which will impact on NO X and NO 2 emissions from vehicles, feed into the calculations so that f-no 2 varies between years to reflect these changes. Additionally the impact of different scenarios on f-no 2 can be modelled using the emissions based calculations of f-no 2 (see section 8 for details of this). The disadvantage of using an emissions based approach to calculate f-no 2 lies in the fact that the Netcen Primary NO 2 model operates at a local scale for specific roadside locations, while the f-no 2 calculated using an emissions based approach here is a national scale average for urban roads. Many AEA Energy & Environment 64

76 of the roads considered in this analysis will have very different characteristics in terms and vehicle fleet composition to the national average so the emissions based f-no 2 may not be representative of all the roads we are actually modelling. To investigate the potential impact of this issue on the modelling results two sets of analysis have been carried out. Firstly, measured annual mean roadside NO 2 concentrations for 24 have been compared with modelled annual mean roadside NO 2 concentrations for 24 using emissions based f- NO 2 in for the base year (see section 6.3). Second, a sensitivity analysis of the impact on the model results of using the local f-no 2 derived using the analysis module of the Netcen Primary NO 2 model for each site has been carried out (see section 7.2 for details of this) NO X Concentrations Projected NO X concentrations at the roadside monitoring site and its paired background site are required. For each country a roadside NO X scaling factor and a background NO X scaling factor has been calculated which can then be applied to the 24 hourly NO X monitoring data to calculate projected hourly concentrations for 21, 215 and 22. Roadside NO X concentrations scaling factors for each country considered are calculated using the ratio of the NO X emissions projections total for urban roads for 21, 215 and 22 to the 25 emissions total for urban roads. Urban roads emission totals have been used in favour of using the national emissions total because all the roadside monitoring sites included in this analysis are located on urban roads. Background NO X concentrations scaling factors for each country have been calculated using a combination of NO X emissions data from GAINS (26) and road transport NO X emissions totals calculated for this analysis (see section 4). To do this the total emissions for all non-road transport sectors from GAINS have been added to our road transport emissions total for all roads. Background scaling factors for each country for 21, 215 and 22 have then been calculated from the ratio of the 21, 215 and 22 emissions total to the 25 emissions total respectively for each country Ozone and Background NO 2 Concentrations There is a certain amount of oxidant in the urban atmosphere. As NO X emissions decrease, partitioning of this oxidant between ozone and NO 2 changes. Generally NO 2 concentrations at background locations are decreasing, but not as steeply as NO X concentrations. Therefore to compensate for this decrease in NO 2, ozone concentrations are generally increasing in urban areas. To calculate ozone and background NO 2 for future years, the oxidant partitioning model (Jenkin, 24) has been used. To use this model to project into the future an increase in regional oxidant of.1 ppb yr -1 has been assumed here based on the analysis presented by Derwent et al (25) of measurements at Mace Head in the west of Ireland. To test the impact of this assumption on modelled ambient NO 2 concentrations in future years, sensitivity analysis on the baseline has been carried out (see section 7.3 for further details of this). A specified local oxidant of.1 (or 1%) has been assumed to apply for all years in the oxidant partitioning model. This number is the proportion of NO X as primary NO 2 for background locations. Analysis presented by AQEG (26) indicates that the changes in f-no 2 for traffic emissions have not had a large impact on NO 2 concentrations away from the roadside, where the dominant contribution is from secondary NO 2. The modelling of ozone and background NO 2 concentrations for future years uses base year meteorological data and base year measured ozone and NO 2 concentrations at the roadside site and paired background site. Using 24 measurement data and meteorological data to project future pollutant concentrations clearly has significant drawbacks. However, as we can t know what the hourly meteorology data will be in the future, this approach has to suffice. AEA Energy & Environment 65

77 6.3 Projection Module Verification Two approaches to verification have been carried out using the projection module of the Netcen Primary NO 2 model set to run for the base year, 24. These are model runs of the projection module using: Our best estimate of local f-no 2 for 24 for each site. This is derived from the analysis module of the Netcen Primary NO 2 model Our baseline for 24 using national scale emissions based calculations of f-no 2 for each country considered in this analysis. The first of these runs is important because it demonstrates the extent to which the model is producing sensible output that agrees with measured concentrations when we have given the model the best estimate of f-no 2 for the local conditions of the modelling site. This run therefore demonstrates how well the model can work in the case study locations given the best possible input data for 24. If the model doesn t work well for these input data, it is likely to perform poorly for other model runs as well. Our second approach to verification is important because it demonstrates how well the model is likely to work at each of the case study sites when using the national scale emissions based calculation of f- NO 2 for urban roads. This is the method for generating f-no 2 numbers for the baseline and scenario results presented in sections 6.4 and 8 respectively. For 24 we can test this using measurement data. It is likely that if the modelled data for 24 significantly over or under predicts the measured data this will also be the case for future years. This second approach is therefore used to test the extent to which the baseline model runs (and therefore scenarios) are capable of representing annual mean NO 2 concentrations for the base year. Because the modelling in this section is for the 24 base year only, scaling factors applied to the input measurement data for these model runs have been set to 1 (i.e. we haven t scaled the input measurement data) Run 1: Using our best estimate of local f-no 2 for 24 Table 6.1 presents comparison statistics between modelled and measured NO 2 concentrations for 24 for the run using our best local estimate of f-no 2 at each roadside site. We have compared the modelled and measured NO 2 concentrations both at an hourly resolution in terms of an R 2 value of hourly measured versus hourly modelled NO 2 concentrations at each site and at an annual mean resolution. A comparison of annual mean measured and modelled concentrations are also presented in Figure 6.1 for all of the sites considered. Table comparison statistics of measured versus modelled NO 2 concentrations using f-no 2 input values derived from the analysis module of the Netcen Primary NO 2 model Country Site Austria Czech Republic Germany R 2 of hourly values Measured annual mean (μg m -3 ) Data Modelled annual capture (%) mean (μg m -3 ) Data capture (%) AT168A AT21A AT38A AT24A CZ66A CZ8A CZ13A CZ11A CZ65A FNV KAS MAV STS AEA Energy & Environment 66

78 Country Site R 2 of hourly values Measured annual mean (μg m -3 ) Data Modelled annual capture (%) mean (μg m -3 ) Data capture (%) GR22A Greece GR4A GR32A GR2A ES1453A Spain ES1262A ES148A ES1438A FI16A Finland FI4A FI6A FI16A FR895A France FR335A FR898A FR91A IT477A_IT IT771A_IT IT116A_IT IT75A_IT Italy 2 IT477A_IT IT771A_IT IT116A_IT IT75A_IT NL248A NL253A NL234A Netherlands NL224A NL244A NL235A Camden Kerbside London A3 Roadside London Cromwell Road UK London Marylebone Road Southwark Roadside Tower Hamlets Roadside For 24 there was insufficient data capture to generate an f-no2 number for this site so we have been unable to run the projection module here using f-no 2 derived the analysis module of the Netcen Primary NO 2 model. 2 For Italy, we have carried out two sets of model runs, one using the backgournd site IT117A in central Milan and one using background site IT123A in a rural location. AEA Energy & Environment 67

79 Figure 6.1. Verification plot of measured vs modelled annual mean concentrations for the 24 base year using locally derived f-no 2 from the analysis module of the Netcen Primary NO 2 model. Modelled annual mean (ugm-3) Sites y = x y = x + 3% y = x - 3% Measured annual mean (ugm-3) In terms of replicating hourly concentrations Table 6.1 shows that in Austria, the Czech Republic, Germany, Finland, France, the Netherlands and the UK, all sites have an R 2 value of greater than.7. This means that at least 7% of the variability in the measured data can be explained by the modelled data and suggests that the model is performing very well in these locations. In Italy and Greece, all but one model run in each location have R 2 values of greater than.7, again suggesting that the model is performing generally well in these locations. In Spain the R 2 values are typically lower, ranging from This suggests that the projection module of the Netcen Primary NO 2 model is not working so well in Spain on an hourly basis. This could be because the background site is known to be a relatively long way from the roadside sites thus increasing the uncertainty in the modelling of ozone concentrations. In terms of reproducing annual mean concentrations, Table 6.1 and Figure 6.1 show that the modelled annual mean at all the sites selected fell within the 3% of the measured annual mean, where 3% is the data quality objective for NO 2 for annual average assessment methods set out in the first daughter directive Run 2: Using emissions based national average f-no 2 for urban roads For the model run using national scale emissions base calculations of f-no 2 for 24 (i.e. the baseline model run for 24), a comparison between measured and modelled hourly and annual mean concentrations is presented in Table 6.2 and Figure 6.2. Table comparison statistics of measured versus modelled NO 2 concentrations using f-no 2 input value derived from the emissions analysis presented in section 4 Country Site Austria R 2 of hourly values Measured annual mean (μ gm -3 ) Data Modelled annual capture (%) mean (μ gm -3 ) Data capture (%) AT168A AT21A AEA Energy & Environment 68

80 Country Site R 2 of hourly values Measured annual mean (μ gm -3 ) Data Modelled annual capture (%) mean (μ gm -3 ) Data capture (%) AT38A AT24A CZ66A CZ8A Czech Republic CZ13A CZ11A CZ65A FNV Germany KAS MAV STS GR22A Greece GR4A GR32A GR2A ES1453A Spain ES1262A ES148A ES1438A FI16A Finland FI4A FI6A FI16A FR895A France FR335A FR898A FR91A IT477A_IT IT771A_IT IT116A_IT IT75A_IT Italy 1 IT477A_IT IT771A_IT IT116A_IT IT75A_IT NL248A NL253A NL234A Netherlands NL224A NL244A NL235A Camden Kerbside London A3 Roadside London Cromwell Road UK London Marylebone Road Southwark Roadside Tower Hamlets Roadside For Italy, we have carried out two sets of model runs, one using the backgournd site IT117A in central Milan and one using background site IT123A in a rural location. AEA Energy & Environment 69

81 Figure 6.2. Verification plot of measured vs modelled annual mean concentrations for the 24 base year using nationwide urban roads emissions based f-no 2 Modelled annual mean (ugm-3) Site y = x y = x + 3% y = x - 3% Measured annual mean (ugm-3) The R 2 values comparing modelled with measured hourly NO 2 for sites modelled are still relatively high for the baseline in 24, which means the model is able to explain much of the variability in the measured data. The relatively high R 2 values suggest that using a different f-no 2 value that is not necessarily representative of local conditions does not impact on the distribution of the hourly results significantly. However, Table 6.2 and Figure 6.2 show that for several sites there is a significant discrepancy between measured and modelled annual mean concentrations. These discrepancies at certain sites suggests that using an f-no 2 that is not necessarily representative of local conditions does impact on the magnitude of the modelled hourly NO 2 concentrations. Sites where the modelled concentrations for the base year are significantly lower than the measured concentrations are the biggest cause for concern in this analysis. This is because in the projections of NO 2 concentrations for future years, we will be likely to similarly under predict concentrations where we have under predicted the base year. Sites where this is likely to be an issue include some of the sites in London and Baden Wűrttemberg. Analysis of the likely impact of under predicting the base year NO 2 concentrations on modelled future concentrations for 21, 215 and 22 is presented in section 7.4 on sensitivity analysis of the model. 6.4 Baseline Modelling Results Annual Mean Concentrations Baseline model projections of annual mean NO 2 concentrations for the roadside sites for 24, 21, 215 and 22 from a 24 base year are presented in Table 6.3 and Figure 6.3. Solid lines in this figure show the baseline projections for each site. Measured annual mean concentrations at these sites for 2-24 are shown using grey lines. Also included are projected concentrations for 21 from relevant Member State s Plans and Programmes reports for 24 to the Commission. This data is only available for monitoring sites that in 24 exceeded the limit value + margin of tolerance. A summary of projected concentrations in 21 at sites for which plans and programmes information is available is given in Table 6.4. AEA Energy & Environment 7

82 At three of the four sites included in the case study for Salzburg and Hallein (Figure 6.3a) the modelled concentration very closely matches the measured concentration in 24. However at AT21A (Zederhaus) the model over predicts in 24 by approximately 1μg m -3. At all four sites the projected annual mean NO 2 concentrations at the roadside are expected to decline between 24 and 22. However, the rate of decline is not predicted to be sufficient to bring the NO 2 concentration at AT38A (Salzburg Rudolfspaltz) or AT24A (Hallein Hagerkreuzung) below the 21 annual mean limit value even by 22. The rate of decrease of NO 2 concentrations at both these sites is predicted to increase throughout the time period covered by the model projections. The Plans and Programmes projection for 21 for annual mean NO 2 concentrations at AT38A provided by Austria (taken from the 23 Plans and Programmes) shows concentrations of a similar order of magnitude to our model projections for this site. For all five sites in Prague considered in this analysis (Figure 6.3b) there is predicted to be an increase in annual mean NO 2 concentrations between 24 and 21. Concentrations are expected to peak in 21 and then decline. By 22 concentrations are predicted to be between 3 and 6 μg m -3 less than in 24. Our model results suggest that the increase in concentrations between 25 and 21 may be sufficient to push the annual mean NO 2 concentration at CZ65A (Pha5-Smichov) marginally above the annual mean limit value for 21 only. The only other site from the Czech Republic case study where the annual mean limit value is predicted to be exceeded is CZ66A (Pha2-Legerova). The limit value at this site is not expected to be met even by 22. In 24 the modelled annual mean NO 2 concentration under predicts the measured by 18.5μg m -3 at CZ66A so annual mean concentrations at this site may be significantly higher into the future as well. For the Finland case study (Figure 6.3c), the model projections show that no sites are expected to exceed the annual mean limit value by 21. This is the case even though the model over predicts concentrations in 24 at all sites considered. It is likely that if the model over predicts in 24, it will continue to do so into the future. NO 2 concentrations are predicted to decline at all four sites between 24 and 22. At the case study sites in Paris (Figure 6.3d) the model over predicts by up to 23.4μg m -3 relative to the measured annual mean concentration in 24. Concentrations are predicted to peak in 21 with 215 concentrations being roughly equal to those in 24. Concentrations are then expected to decrease rapidly between 215 and 22. At all the sites considered the annual mean limit value is predicted to be exceeded for the entire period over which projections have been calculated. The Plans and Programmes data presented for the French sites considered in our analysis all significantly under predict compared with our model projections. This may be partly because our model is over predicting in the base year. However the difference in concentrations between our model projections and the Plans and Programmes projections cannot be entirely accounted for by this since our model suggests concentrations will rise between 24 and 21 and the Plans and Programmes data point to a decrease in concentrations over the same period. Both sets of projections suggest that the annual mean limit value will not be achieved in 21. At the sites included in the Athens case study (Figure 6.3e) concentrations are predicted to decrease over the time period covered by our projections. The higher the initial concentrations, the greater the decrease is expected to be. However, the decrease in concentrations is not predicted to be sufficient to bring concentrations at GR32A (PATISION) down to the limit value even by 22. At GR2A (ATHINAS), the limit value is also predicted to be exceeded in 21. At all four sites considered in this case study, the model under predicts relative to measured concentrations in 24 by up to 21μg m -3. The model projections for 21, 215 and 22 therefore are also likely to under predict suggesting that for at least two of these sites the limit value may be difficult to achieve. Within the Baden Wűrttemberg case study (Figure 6.3f) the model projections suggest that there will be general decline in concentrations between 24 and 22. At three of the four sites modelled, the model under predicts measured concentrations in 24. At STS (Stuttgart Strassenstation) this under prediction is particularly significant (16.8g m -3 ) suggesting that concentrations in 21 may be nearer to the value suggested in the Plans and Programmes for this site. The model results suggest that at STS the limit value will be exceeded the entire period 21-22, with exceedences in 21 also likely at KAS (Karlsruhe Strassenstation) and MAV (Manheim Strassenstation). AEA Energy & Environment 71

83 For the Italian case study sites in Milan (Figure 6.3g) two sets of model runs have been carried out using alternative background monitoring stations IT117A (P.CO LAMBRO 3153) and IT123A (ARCONATE 3154). This is because in section 3 we found that the Netcen Primary NO 2 model was sensitive to the choice of background site in Milan. Both sets of model runs under predict the measure concentration in 24 at all sites considered. This under prediction is greatest for the model runs using IT123A, which is a rural site outside of Milan. Both sets of model runs predict that concentrations will decrease over time. For the runs using IT117A, the concentrations at all four roadside sites are predicted to remain above the annual mean limit value for the entire period under consideration. For the runs using IT123A, concentrations at two of the sites, IT116A (SENATO MARINA 31537) and IT75A (VERZIERE 3154) are projected to decrease below the limit value by 22. In the Netherlands case study (Figure 6.3h) the model under predicts the measured annual mean concentrations in 24 at each site by between 2.2 and 11.2μg m -3. The model projections suggest that for all six sites considered concentrations will decrease between 24 and 22. However at NL253A (Den Haag-Veerkade) this decrease is predicted to be insufficient to meet the annual mean limit value before 22. At NL248 (Breukelen-Snelweg) the baseline model projections suggest that the annual mean limit value won t be met until after the 21 compliance deadline. For the sites in the Barcelona case study (Figure 6.3i), the model under predicts annual mean concentrations in 24 relative to measured concentrations except at ES1262A (ES1262A-AD- SABADELL) where the model over predicts by 6μg m -3. At all four sites included in this case study, there is predicted to be little change in concentrations between 24 and 21. Concentrations are predicted to decrease from 21 onwards. The greatest rate of decrease is predicted to occur between 215 and 22. At ES1262A (ES1262A-AD-SABADELL), ES148A (ES148A-IJ-GRACIA- SANT GERVASI) and ES1438A (ES1438A-IH-BARCELONA(Eixample)) the lack of change in concentrations predicted between 25 and 21 means that these sites are expected to exceed the limit value in 21 The Plans and Programmes information on predicted annual mean concentrations for 21 at ES148A and ES1438 gives higher concentrations for this year than our model results suggest. However, a significant part of this discrepancy can probably be accounted for by the under prediction in 24 compared to measured concentrations at these sites in our model results. If our results for 21 were increased by the difference between the modelled and measured concentrations in 24 they would show good agreement with the Plans and Programmes results. At the UK case study sites in London the annual mean NO 2 limit value is not predicted to be achieved at any site before 22. This is the case despite the fact that there is predicted to be a relatively constant decline in NO 2 annual mean concentrations between 24 and 22. At London Marylebone Road (GB682A) the model significantly under predicts the measured concentrations in 24. All model projections fall within the range given in Plans and Programmes for the London urban roads exceedence situation. It is likely that differences between local conditions and the national average for urban roads cause the discrepancies in Figure 6.3 found at many sites between our baseline model prediction in 24 and measurements for this year. These local differences may result in f-no 2 that is much higher or lower than is actually found at each site hence causing the concentrations to be over or under predicted. AEA Energy & Environment 72

84 Figure 6.3. Baseline projections for 21, 215 and 22 (solid coloured lines). Measured concentrations for 2-24 are show using grey lines. Projected concentrations from Plans and Programmes for member states (where relevant) are shown in as coloured points with bars attached to show the possible range of values a) Austria b) Czech Republic 6 8 Modelled annual mean NO2 concentration (ugm-3) LV + MOT AT168A AT21A AT38A AT24A Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 CZ66A CZ8A CZ13A 1 CZ11A CZ65A c) Finland d) France 6 12 Modelled annual mean NO2 concentration (ugm-3) LV + MOT FI16A FI4A FI6A FI16A Modelled annual mean NO2 concentration (ugm-3) LV + MOT FR895A FR335A FR898A FR91A e) Greece f) Germany AEA Energy & Environment 73

85 1 1 LV + MOT LV + MOT Modelled annual mean NO2 concentration (ugm-3) GR22A GR4A GR32A GR2A Modelled annual mean NO2 concentration (ugm-3) FNV KAS MAV STS Figure 6.3. continued g) Italy Modelled annual mean NO2 concentration (ugm-3) LV + MOT IT477A_IT117 IT771A_IT117 IT116A_IT117 IT75A_IT117 Modelled annual mean NO2 concentration (ugm- 3) LV + MOT IT477A_IT123 IT771A_IT123 IT116A_IT123 IT75A_IT h) Netherlands i) Spain 7 7 Modelled annual mean NO2 concentration (ugm-3) LV + MOT NL248A NL253A NL234A NL224A NL244A NL235A Modelled annual mean NO2 concentration (ugm- 3) LV + MOT ES1453A ES1262A ES148A ES1438A j) UK AEA Energy & Environment 74

86 11 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT Camden Kerbside London A3 Roadside London Cromwell Road 2 London Marylebone Road Southwark Roadside Tower Hamlets Roadside Table 6.3. Baseline projections of annual mean roadside NO 2 concentrations for 24, 21, 215 and 22 from a 24 base Country Site Modelled annual mean NO 2 concentration (μg m -3 ) Percentage of total hours per year modelled (%) AT168A Austria Czech Republic Germany Greece Spain Finland France Italy 1 AT21A AT38A AT24A CZ66A CZ8A CZ13A CZ11A CZ65A FNV KAS MAV STS GR22A GR4A GR32A GR2A ES1453A ES1262A ES148A ES1438A FI16A FI4A FI6A FI16A FR895A FR335A FR898A FR91A IT477A_IT IT771A_IT IT116A_IT AEA Energy & Environment 75

87 Country Site Modelled annual mean NO 2 concentration (μg m -3 ) Percentage of total hours per year modelled (%) IT75A_IT IT477A_IT IT771A_IT IT116A_IT IT75A_IT NL248A NL253A NL234A Netherlands NL224A UK NL244A NL235A Camden Kerbside London A3 Roadside London Cromwell Road London Marylebone Road Southwark Roadside Tower Hamlets Roadside For Italy, we have carried out two sets of model runs, one using the background site IT117A in central Milan and one using background site IT123A in a rural location. Table 6.4. Projected NO 2 annual mean concentrations for 21 from 24 Plans and Programmes. (23 for Austria as this was the most recent data available) Country Site AT168A Range of Projected concentrations for 21 (μg m -3 ) Regional background Background Total No exceedence of LV + MOT Austria AT21A No exceedence of LV + MOT AT38A AT24A CZ66A CZ8A No exceedence of LV + MOT Czech Republic CZ13A No exceedence of LV + MOT CZ11A No exceedence of LV + MOT CZ65A No exceedence of LV + MOT FI16A No exceedence of LV + MOT Finland FI4A No exceedence of LV + MOT FI6A No exceedence of LV + MOT FI16A No exceedence of LV + MOT FR895A 1 µg/m3 35 µg/m3 (Average of Paris and departments 92, 93, 94) 79-8ug/m3 36 µg/m3 (Average of Paris France FR335A 1 µg/m3 and departments 92, 93, 94) 71ug/m3 37 µg/m3 (Average of Paris FR898A 1 µg/m3 and departments 92, 93, 94) 68-7ug/m3 FR91A 1 µg/m3 38 µg/m3 (Average of Paris and departments 92, 93, 94) 64-68ug/m3 FNV No exceedence of LV + MOT Germany KAS MAV STS GR22A No data available Greece GR4A No data available GR32A No data available GR2A No data available Italy IT477A No data available AEA Energy & Environment 76

88 Range of Projected concentrations for 21 (μg m -3 ) IT771A No data available IT116A No data available IT75A No data available NL248A No exceedence of LV + MOT NL253A NL234A Netherlands NL224A NL244A NL235A ES1453A No exceedence of LV + MOT Spain ES1262A No exceedence of LV + MOT ES148A ES1438A GB636A No exceedence of LV + MOT GB659A UK GB695A GB682A GB667A GB624A Exceedences of the Hourly NO 2 Limit Value The annual mean limit value is expected to be the most stringent in 21. Therefore we have focussed most of our analysis on this metric. However, it is possible that the relative stringency of the annual mean and hourly limit values could be affected by changes in f-no 2. Therefore we have also calculated the number of hours exceeding the hourly limit value (2μg m -3 ). Table 6.5 presents this data. Where the number of exceedences is in excess of the 18 allowed the numbers are presented in bold. Table 6.5. Number of exceedences of the hourly limit value (2μg m -3, 18 exceedences allowed per year) Country Site Measured number of hours exceeding NO 2 LV Modelled number of hours exceeding NO 2 LV Plans and Programmes AT168A 1 2 No exceedence Austria Czech Republic Finland France AT21A No exceedence AT38A 2 No exceedence AT24A 1 No exceedence CZ66A No exceedence CZ8A No exceedence CZ13A No exceedence CZ11A 9 No exceedence CZ65A No exceedence FI16A No exceedence FI4A No exceedence FI6A No exceedence FI16A 1 No exceedence FR895A No exceedence FR335A No exceedence FR898A No exceedence AEA Energy & Environment 77

89 Country Site Measured number of hours exceeding NO 2 LV Modelled number of hours exceeding NO 2 LV Plans and Programmes FR91A No exceedence Greece Germany Italy GR22A No data available GR4A No data available GR32A No data available GR2A No data available FNV No exceedence KAS No exceedence MAV 1 22 No exceedence STS No exceedence IT477A_IT No data available IT771A_IT No data available IT116A_IT No data available IT75A_IT No data available IT477A_IT No data available IT771A_IT No data available IT116A_IT No data available IT75A_IT No data available NL248A No exceedence NL253A No exceedence Netherlands NL234A 3 2 No exceedence NL224A No exceedence Spain UK NL244A 1 1 No exceedence NL235A No exceedence ES1453A 2 2 No exceedence ES1262A No exceedence ES148A No exceedence ES1438A No exceedence Camden Kerbside 2 6 No exceedence London A3 Roadside No exceedence London Cromwell Road No exceedence London Marylebone Road Southwark Roadside 2 No exceedence Tower Hamlets Roadside No exceedence The data presented in Table 6.5 show that modelled exceedences of the hourly limit value are predicted to occur at all the sites included in the French case study and at IT477A (MARCHE 31526) using the urban background site IT117A (P.CO LAMBRO 3153) in central Milan. At all the other sites included in this analysis, our baseline projections suggest that the hourly limit value will be achieved by 21. However, at some sites, the modelled number of exceedences in 24 significantly under predicts the measured number of exceedences. This is most noticeable at CZ66A (Pha2-Legerova) and London Marylebone Road (GB682A). This under prediction is likely to be caused by the differences between f-no 2 calculated as a national average for urban roads (used in our baseline emissions projections) and the actual local f-no 2 at each of these sites. Further work investigating the impact of using a national average factor rather than a locally derived f-no 2 on numbers of hourly exceedences is presented in section 7.4. Despite the modelled under prediction at a few sites, the modelling results in Table 6.5 suggest that there will be relatively few sites where the hourly limit value is unlikely to be met in 21. At sites where the hourly limit value is not expected to be met, the annual mean limit value is also predicted to AEA Energy & Environment 78

90 be exceeded. This suggests that the annual mean limit value will continue to be more stringent into the future as there are many sites where the model predicts that the annual mean limit value will be exceeded where there are no exceedences of the hourly limit value. Therefore for the remaining analysis presented in this report the focus will be on the annual mean limit value. 6.5 Conclusions The model verification presented for 24 (i.e. the 24 modelled annual mean compared with the measured annual mean) showed that there was good agreement with all sites considered in this analysis falling within the 3% data quality objectives set out in the first daughter directive. In terms of a comparison of modelled and measured hourly data in 24 all sites had an R 2 value of greater than.7 except for a few sites in Spain, Italy and Greece. Baseline annual mean NO 2 concentrations are predicted to decline between 25 and 22 at most of the case study locations. Little or no decline is predicted between 25 and 21 at the case study sites in the Czech Republic and Spain, presumably due to a combination of relatively modest decreases in NO X emissions from traffic and an increase in f-no 2 at these locations. An increase in annual mean NO 2 concentration between 25 and 21 is predicted at the case study sites in France, due to the high roadside NO X concentrations at these sites and high predicted f-no 2 in 21. In all other locations the predicted increase in f-no 2 is not large enough to offset the decrease in NO X emissions. At many of the sites considered the annual mean NO 2 limit value is predicted to be exceeded in 21. At a significant number of these sites the annual mean concentration is expected to remain above the limit value at least until 22. The baseline model results suggest that the annual mean limit value will continue to be the more stringent NO 2 limit value up to and including 22. Therefore the focus of the remaining work in this report will be on this limit value. AEA Energy & Environment 79

91 7 Future Ambient NO 2 Concentrations: Baseline Sensitivity Analysis 7.1 Introduction The previous section presents our baseline projections of ambient NO 2 concentrations at the roadside sites selected for analysis in this report. This section presents further analysis of the baseline projections in terms of sensitivity of the model to changes in input data. The following sensitivity analysis is presented: Sensitivity of modelled NO 2 concentrations to changes in f-no 2 Sensitivity of modelled NO 2 concentrations to changes in future regional oxidant levels Analysis of how under prediction of the base year measured annual mean concentrations using the baseline projections may impact on future concentrations. Conclusions are then drawn regarding the impact of the assumptions used in the baseline modelling on modelled concentrations of NO 2 at the roadside. 7.2 Model Sensitivity to f-no 2 As explained in section 6, the baseline projections use f-no 2 calculated using an emissions based approach. This is a national scale average for urban roads which may or may not represent local conditions adequately. The work presented in this section aims to investigate the impact of applying national average factors rather than local factors on the model results. Three model runs have been carried out to assess this. These are model runs using: Our baseline where f-no 2 is derived from national scale emissions based calculations of f-no 2 for urban roads for each country considered in this analysis. Our best estimate of local f-no 2 for 24 derived from the analysis module of the Netcen Primary NO 2 model. f-no 2 for future years has been kept constant at 24 levels for this run Baseline emissions calculated f-no 2 for urban roads kept constant for 21, 215 and 22 at 25 value. Results of this analysis are presented in Figure 7.1. A comparison of the modelled baseline where temporally variable emissions based estimate f-no 2 have been used (solid coloured line) and the sensitivity test where we have kept 25 emissions base f-no 2 constant into the future (dotted coloured line) shows how sensitive the model is to f-no 2 changes into the future. These two sets of projections will always start at the same concentration in 24 so any difference in 21 is purely down to the difference in f-no 2, not differences in where the projections start. At most sites, Figure 7.1 shows that the baseline predicts higher concentrations into the future than this sensitivity test. This results from the fact that for most countries f-no 2 increases into the future in the baseline projections (see Figure 4.4), whereas in the sensitivity tests presented here f-no 2 remains constant. Generally at sites with higher total NO 2 concentrations the difference between these two runs is much greater than at sites with less high annual mean NO 2 concentrations. This probably results from a larger NO X roadside increment at sites with relatively high concentrations. Where there is a large NO X roadside increment the impact of higher future f-no 2 will be greater than where there is a lower NO X roadside increment, as there will be more NO X emissions from vehicles so the absolute amount of direct NO 2 emissions will be larger. The contribution from secondary NO 2 is greater at sites with lower relative concentrations. The second sensitivity test presented in Figure 7.1 uses a constant locally derived f-no 2 (dashed coloured line). This is useful to compare with the sensitivity test where f-no 2 derived from emissions analysis has been kept constant into the future (dotted coloured line). This is because it gives an indication of the magnitude of the impact of using a national average f-no 2 for urban roads on the AEA Energy & Environment 8

92 model projections compared with using a locally derived factor. At many sites the impact is relatively small. There are two possible reasons for this: The NO X roadside increment is small enough at these sites that varying f-no 2 has little impact on the NO 2 concentration The national urban roads f-no 2 for these sites is similar to the local f-no 2 At several of the sites selected for analysis with the largest measured annual mean NO 2 concentrations the discrepancy between these two lines is relatively large. This suggests that the f-no 2 at these sites is poorly represented by the national average. This can be for a selection of reasons including the typical traffic flow along the road adjacent to the site not reflecting the total vehicle fleet for that country. For example at London Marylebone road (GB682A), the introduction of the congestion charge altering traffic flows and the introduction of more buses with fitted particulate traps is thought to have increased f-no 2 significantly above the national average for urban roads. AEA Energy & Environment 81

93 Figure 7.1. Graphs showing sensitivity of the baseline to using different f-no 2 input data for 21, 215 and 22. Grey lines show measured annual mean; solid coloured lines represent the baseline; dashed lines are the results with constant f-no 2 between 25 and 22, as derived from 24 monitoring data; dotted lines show the results with constant f-no 2 between 25 and 22, as derived for 25 from the emission inventory calculations. a) Austria b) Czech Republic 6 8 Modelled annual mean NO2 concentration (ugm-3) LV + MOT AT168A AT21A AT38A AT24A Modelled annual mean NO2 concentration (ugm-3) LV + MOT CZ66A CZ8A CZ13A CZ11A CZ65A c) Finland d) France 6 12 Modelled annual mean NO2 concentration (ugm-3) LV + MOT FI16A FI4A FI6A FI16A Modelled annual mean NO2 concentration (ugm-3) LV + MOT FR895A FR335A FR898A FR91A e) Greece f) Germany AEA Energy & Environment 82

94 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT GR22A GR4A GR32A GR2A Modelled annual mean NO2 concentration (ugm-3) LV + MOT FNV KAS MA V STS Figure 7.1. continued g) Italy Modelled annual mean NO2 concentration (ugm-3) LV + MOT IT477A_IT117 IT771A_IT117 IT116A_IT117 IT75A_IT117 Modelled annual mean NO2 concentration (ugm- 3) LV + MOT IT477A_IT123 IT771A_IT123 IT116A_IT123 IT75A_IT h) Netherlands i) Spain 7 7 Modelled annual mean NO2 concentration (ugm-3) LV + MOT NL248A NL253A NL234A NL224A NL244A NL235A Modelled annual mean NO2 concentration (ugm- 3) LV + MOT ES1453A ES1262A ES148A ES1438A j) UK AEA Energy & Environment 83

95 11 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT Camden Kerbside London A3 Roadside London Cromw ell Road 2 London Marylebone Road Southw ark Roadside Tow er Hamlets Roadside Model Sensitivity to Future Regional Oxidant Levels For the baseline modelling, regional oxidant levels have been assumed to increase by.1 ppb yr -1. However, changes in regional oxidant for future years are notoriously difficult to predict. Therefore, in order to assess the impact of this assumption on modelled NO 2 concentrations for future years, two extra model runs have been done using baseline conditions, but with regional oxidant levels remaining at 24 levels (i.e. a change of ppb yr -1 ) and with regional oxidant levels increasing at twice the rate assumed in the baseline (i.e. an increase of.2 ppb yr -1 ). Table 7.1 shows the input regional oxidant levels used in this analysis. The modelling results are presented in Figure 7.2. Table 7.1. Regional oxidant levels used in the model runs testing the sensitivity of the baseline to variations in regional oxidant trends Assumed increase per year (ppb yr -1 ) Country (background station assumed to represent regional conditions) Total regional oxidant (ppb yr -1 ) (baseline conditions) UK (London Teddington) France (FR918A) Germany (FMI) Spain (Sant Cugat del Vallès) Greece (GR35A) Italy (IT123A) Czech Republic (CZ2A) Austria (AT75A) Netherlands (NL25A) (no increase) Finland (FI124A) UK (London Teddington) France (FR918A) Germany (FMI) Spain (Sant Cugat del Vallès) Greece (GR35A) Italy (IT123A) Czech Republic (CZ2A) Austria (AT75A) Netherlands (NL25A) Finland (FI124A) AEA Energy & Environment 84

96 Assumed increase per year (ppb yr -1 ).2 (double the increase in the baseline) Country (background station assumed to represent regional conditions) Total regional oxidant (ppb yr -1 ) UK (London Teddington) France (FR918A) Germany (FMI) Spain (Sant Cugat del Vallès) Greece (GR35A) Italy (IT123A) Czech Republic (CZ2A) Austria (AT75A) Netherlands (NL25A) Finland (FI124A) For all the case study sites in all ten countries considered a very consistent pattern of the impact of future regional oxidant levels on future NO 2 concentrations is evident in Figure 7.2. Where regional oxidant is assumed to remain constant for future years (dashed lines on Figure 7.2) modelled NO 2 concentrations are slightly lower than for the baseline (solid lines on Figure 7.2). Where regional oxidant is assumed to increase at twice the rate of the baseline, i.e..2 ppb yr -1 increase (dotted lines on Figure 7.2) concentrations are slightly higher than the baseline. The magnitude of the difference between the two sensitivity runs and the baseline increases the further forward into the future we go. This is because the difference in the regional oxidant for the different runs, and the uncertainty in the assumed regional oxidant increases each year. By 22, the baseline typically has annual mean NO 2 concentrations of 2-3 μg m -3 greater than for ppb yr -1 increase run and 2-3 μg m -3 less than the.2 ppb yr -1 increase run. Thus the assumed change in regional oxidant has a noticeable and consistent influence on the predictions of annual mean NO 2 at the roadside. The magnitude of this influence is, however, smaller than that of changes in f-no 2, particularly at the sites with the highest measured concentrations. AEA Energy & Environment 85

97 Figure 7.2. Graphs showing sensitivity of the baseline to changes in regional oxidant levels for 21, 215 and 22. Grey lines show measured annual mean; solid coloured lines represent the baseline; dashed lines are the run with ppb yr -1 change in regional oxidant; dotted lines show the run for.2 ppb yr -1 increase in regional oxidant. a) Austria b) Czech Republic 6 8 Modelled annual mean NO2 concentration (ugm-3) LV + MOT AT168A AT21A AT38A AT24A Modelled annual mean NO2 concentration (ugm-3) LV + MOT CZ66A CZ8A CZ13A CZ11A CZ65A c) Finland d) France 6 14 Modelled annual mean NO2 concentration (ugm-3) LV + MOT FI16A FI4A FI6A FI16A Modelled annual mean NO2 concentration (ugm-3) LV + MOT FR895A FR335A FR898A FR91A e) Greece f) Germany AEA Energy & Environment 86

98 1 9 LV + MOT 8 Modelled annual mean NO2 concentration (ugm-3) GR22A GR4A GR32A GR2A Modelled annual mean NO2 concentration (ugm-3) LV + MOT FNV KAS MA V STS Figure 7.2. continued g) Italy Modelled annual mean NO2 concentration (ugm-3) LV + MOT IT477A_IT117 IT771A_IT117 IT116A_IT117 IT75A_IT117 Modelled annual mean NO2 concentration (ugm-3) LV + MOT IT477A_IT123 IT771A_IT123 IT116A_IT123 IT75A_IT h) Netherlands i) Spain AEA Energy & Environment 87

99 7 7 Modelled annual mean NO2 concentration (ugm-3) LV + MOT NL248A NL253A NL234A NL224A NL244A NL235A Modelled annual mean NO2 concentration (ugm-3) LV + MOT ES1453A ES1262A ES148A ES1438A j) UK 11 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT Camden Kerbside 3 London A3 Roadside 2 London Cromw ell Road 2 London Marylebone Road 1 Southw ark Roadside Tow er Hamlets Roadside The Impact of Under Predicting Base Model Results on Modelled Future NO 2 Concentrations Comparison of modelled and measured NO 2 concentrations for 24 base year showed that for some sites the model significantly over or under predicted concentrations because a national average f-no 2 have been used rather than a local one (see section 6.3). This is most problematic where measured 24 data is very high compared with the modelled data in 24 because it is likely in this instance that our projections to 21, 215 and 22 will also under-predict concentrations. This could result in incorrect conclusions being drawn that the limit value is likely to be complied with earlier than it actually will be in reality. To investigate the potential impact of this, three of the sites where the modelled results under predict the measured results in 24 (both annual mean and for the hourly limit value) have been selected for further analysis. These are CZ66A (Pha2-Legerova), STS (Stuttgart Strassenstation) and London Marylebone Road (GB682A). For the base year at these sites the model run using f-no 2 derived from the analysis module of the Netcen Primary NO 2 model produced concentrations that better matched the measured NO 2 concentrations. Therefore we have started in 24 with this as the base year f-no 2 for these sites. f-no 2 for urban roads for future years has then been calculated based on the trends in the emissions based calculations of f-no 2 for future years. However, because the f-no 2 AEA Energy & Environment 88

100 at these sites in 24 derived from module 1 of the Netcen Primary NO 2 model is much higher than f- NO 2 derived using the emissions based approach, f-no 2 is unlikely to increase at the rate suggested in the emissions calculations. Therefore we have assumed that f-no 2 will increase at half of this rate. This is clearly a generalisation of what might actually happen. However, the assumption will give some indication of how much using the emissions based f-no 2 is liable to cause the model to under predict concentrations in future years. The f-no 2 values used in these model runs are presented in Table 7.2 Table 7.2. f-no 2 values used to test the impact of the emissions based f-no 2 for urban roads underpredicting roadside annual mean NO 2 concentrations at certain sites Baseline Sensitivity test f-no 2 value used (%) Country Site Czech Republic CZ66A Germany STS UK London Marylebone Road Czech Republic CZ66A Germany STS UK London Marylebone Road Annual mean concentrations The results of this model run in terms of annual mean concentrations for these three sites are presented in Figure 7.3. The baseline is also shown on these plots for comparison purposes. The NO 2 concentration in 24 for the sensitivity test is significantly higher than for the baseline at all three sites considered. This difference then narrows at all three sites such that by 215 the annual mean NO 2 concentration projections for the sensitivity test and the baseline are very similar at both STS (Stuttgart Strassenstation) and London Marylebone Road (GB682A). The difference at CZ66A (Pha2-Legerova) narrows to approximately 5μg m -3 by 215 and 22 with the sensitivity test remaining the higher modelled concentration. This analysis therefore suggests that under predicting the f-no 2 by using a national average for urban roads rather than a locally derived factor will have a significant impact on annual mean concentrations in 24, but this impact will become less and less significant the further into the future the projections are made. AEA Energy & Environment 89

101 Figure 7.3. Graphs showing the impact of the emissions based f-no 2 for urban roads under-predicting roadside annual mean NO 2 concentrations at certain sites in 21, 215 and 22. a) CZ66A (Pha2-Legerova) b) STS (Stuttgart Strassenstation) 8 8 Modelled annual mean NO2 concentration (ugm-3) LV + MOT CZ66A Measured CZ66A Baseline CZ66A Sensitivity test Modelled annual mean NO2 concentration (ugm-3) LV + MOT STS Measured STS Baseline STS Sensitivity test c) London Marylebone Road (GB682A) 11 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT London Marylebone Road Measured London Marylebone Road Baseline London Marylebone Road Sensitivity test Comparison with the hourly mean limit value Table 7.3 presents a comparison of this sensitivity test with the baseline in terms of exceedences of the hourly mean limit value. At both CZ66A (Pha2-Legerova) and London Marylebone Road (GB682A) the sensitivity test using a locally derived f-no 2 in 24 predicts a significantly greater number of exceedences than the baseline. Although this sensitivity test still under predicts relative to the measurements in 24, it is much closer than the baseline to the measured values. After 24 at both sites the sensitivity test predicts that the number of exceedences will decrease rapidly such that by 22 there will be no exceedences at London Marylebone road and only 24 at CZ66A. However in 21 the sensitivity test suggests that the hourly limit value will be exceeded at both these sites. AEA Energy & Environment 9

102 At STS (Stuttgart Strassenstation) the sensitivity test predicted the same number of exceedences in 24 to the number that were measured. This suggests that at this site the locally derived f-no 2 enabled the model to perform better than the baseline. Table 7.3. Number of exceedences of the hourly limit value (2μg m -3, 18 exceedences allowed per year) Number of hours exceeding NO2 LV Country Site Czech Republic Germany UK CZ66A STS London Marylebone Road Measured Baseline Sensitivity test Measured Baseline Sensitivity test Measured Baseline Sensitivity test Relative stringency of the NO 2 limit values At all three sites considered above, using local f-no 2 estimates, the annual mean limit value appears to be more difficult to meet than the hourly limit value. This finding re-enforces that the focus of this report should be on the achievability of the annual mean limit value as there are unlikely to be situations where the annual mean limit value has been met and the hourly limit value has not. 7.5 Conclusions of Sensitivity analysis The sensitivity analysis presented in this section suggests that the model is generally very sensitive to the f-no 2 used. It is therefore important to produce high quality projections of future f-no 2 when modelling ambient roadside NO 2 concentrations for 21, 215 and 22. This supports the approach we have adopted in this report of using an emissions based approach to project f-no 2 as it is not possible to project f-no 2 into the future base on current local monitoring data. However, the relatively high sensitivity of the model to the f-no 2 input data does cause problems in terms of our adopted approach because the emissions numbers we have used to generate the f-no 2 projections are a national average and therefore not necessarily representative of any one given monitoring site. In order to produce local emissions-based f-no 2 estimates a lot of further work would be required and is beyond the scope of the current study. In future work investigating f-no 2 it would be recommended to use local scale emissions data in generating f-no 2 projections. There are local inventories available, for example the London Atmospheric Emissions Inventory (GLA, 26), which contain data on local vehicle fleets, locally specific vehicle trip characteristics and road networks. The model seems most sensitive to changes in f-no 2 at sites with relatively high NO X concentrations. This is because at these sites there is likely to be more NO X emitted from vehicles so changing the proportion of this NO X modelled as direct NO 2 emissions has a greater impact on the absolute amounts of NO 2 around. In terms of the model sensitivity to changes in levels of future regional oxidant, varying this input parameter does have an impact on modelled ambient NO 2 concentrations. However the model seems to be less sensitive to changes in regional oxidant level than to changes in f-no 2 AEA Energy & Environment 91

103 8 Future Ambient NO 2 Concentrations: Scenario Projections 8.1 Introduction Section 6 presented our baseline model projections of ambient NO 2 concentrations in 21, 215 and 22. Section 7 then gave details of sensitivity analysis carried out to test how important the different input parameters are in our baseline and also to help put the baseline results in context. This section presents scenario modelling used to examine the impact of different options for future emission limits and exhaust after treatment technologies on estimates of ambient NO 2 concentrations. Ambient NO 2 concentrations have been predicted in 21, 215 and 22 for a total of four scenarios in addition to the baseline. The scenario assumptions are listed in Table 8.1. Table 8.1 Summary of NO X emissions and f-no 2 scenarios Scenario Description baseline NO X projections include the impact of Euro 5 and Euro 6 (LDV). Baseline f-no 2 scenario 1 scenario 2 scenario 3 scenario 4 NO X projections include the impact of Euro 5 and Euro 6 (LDV). More pessimistic f-no 2 for HDV than the baseline NO X projections include the impact of Euro 5, Euro 6 (LDV) and an estimate for Euro VI (HDV). More optimistic f-no 2 for HDV than the baseline NO X projections include the impact of Euro 5 and Euro 6 (LDV). More optimistic f-no 2 for LDV at Euro 6 than the baseline NO X projections include the impact of Euro 5, Euro 6 (LDV) and an estimate for Euro VI (HDV). Same f-no 2 and NO X control assumption for HDV as scenario 2 and more optimistic f-no 2 for LDV than scenario Modelling Assumptions and Emissions Projections Figure 8.1 summarises the emission projections for the different scenarios in terms of total NO X and NO 2 emissions and average f-no 2 across the ten member states considered. Figure 8.1 is the same as figure 4.5 and is reproduced here for easy reference. NO X emissions decline steeply to 22 and are lowest for scenarios 2 and 4 in 22. f-no 2 is predicted to increase steeply from 25 to 215 and then starting to decline for scenarios 3 and 4. Thus NO 2 emissions are predicted to increase from 2 to 21 in contrast to the decline in NO X emissions. NO 2 emissions are then predicted to flatten off to 215 and then decline to roughly equivalent to 25 values in 22 for the baseline and scenarios 1 and 2. By 22 the decease in NO X emissions is sufficient to offset the increase in f-no 2. NO 2 emissions are projected to decrease more steeply to 22 for scenarios 3 and 4 to values below 2 emissions as a result of a decrease in f-no 2 combined with NO X emission reductions. Further details of the methods used and country specific input parameters are given in section 4. AEA Energy & Environment 92

104 Figure 8.1 Graphs summarising urban road traffic NO x and NO 2 emissions summed across the ten member states considered along with average urban f-no 2 for the different emission projection scenarios a) NO X emissions (ktonnes per year) b) NO 2 emissions (ktonnes per year) Baseline Scenario 1 Scenario 2 Scenario 3 Scenario Baseline Scenario 1 Scenario 2 Scenario 3 Scenario ktonnes per year ktonnes per year year year c) f-no 2 (percent) 4% 35% 3% Baseline Scenario 1 Scenario 2 Scenario 3 Scenario 4 f-no2 (percent) 25% 2% 15% 1% 5% % year 8.3 Scenario Modelling Results Figure 8.2 presents graphs showing the impact of the scenarios on modelled annual mean NO 2 concentrations in 21, 215 and 22. AEA Energy & Environment 93

105 AEA/ENV/R/244 Issue 2 Figure 8.2. Graphs showing the impact of the scenarios on modelled NO 2 concentrations in 21, 215 and a Austria 6 Measured Baseline Scenario 1 Scenario 2 Modelled annual mean NO2 concentration (ugm-3) LV + MOT AT168A 1 AT21A AT38A AT24A Scenario 3 Scenario 4 AEA Energy & Environment 94

106 AEA/ENV/R/244 Issue 2 8.2b Czech Republic 8 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 CZ66A CZ8A CZ13A 1 CZ11A CZ65A Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 95

107 AEA/ENV/R/244 Issue 2 8.2c Finland Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 1 FI16A FI4A FI6A FI16A Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 96

108 AEA/ENV/R/244 Issue 2 8.2d France 13 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 3 FR895A FR335A 2 FR898A 1 FR91A Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 97

109 AEA/ENV/R/244 Issue 2 8.2e Greece 1 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 GR22A GR4A 1 GR32A GR2A Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 98

110 AEA/ENV/R/244 Issue 2 8.2f Germany 9 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 FNV KAS 1 MA V STS Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 99

111 AEA/ENV/R/244 Issue 2 8.2g Italy (IT117A background site) 9 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 IT477A_IT117 IT771A_IT117 1 IT116A_IT117 IT75A_IT Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 1

112 AEA/ENV/R/244 Issue 2 8.2h Italy (IT123A background site) 9 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 IT477A_IT123 IT771A_IT123 1 IT116A_IT123 IT75A_IT Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 11

113 AEA/ENV/R/244 Issue 2 8.2i Netherlands 7 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT 2 NL248A NL253A NL234A 1 NL224A NL244A NL235A Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 12

114 AEA/ENV/R/244 Issue 2 8.2j Spain 7 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT ES1453A ES1262A 1 ES148A ES1438A Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 13

115 AEA/ENV/R/244 Issue 2 8.2k UK (three of six sites) 11 Measured Baseline Scenario 1 Modelled annual mean NO2 concentration (ugm-3) LV + MOT London Cromw ell Road 2 London Marylebone Road Southw ark Roadside Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 14

116 AEA/ENV/R/244 Issue 2 8.2j UK (remaining sites) Measured Baseline Modelled annual mean NO2 concentration (ugm-3) LV + MOT Camden Kerbside London A3 Roadside Tow er Hamlets Roadside Scenario 1 Scenario 2 Scenario 3 Scenario 4 AEA Energy & Environment 15

117 The predictions of annual mean NO 2 concentration were found to be most sensitive to the scenario assumptions at the sites with the highest NO X concentrations. As expected the predicted NO 2 concentration for scenario 1 are higher than the baseline since f-no 2 for heavy duty vehicles is assumed to be higher than the baseline. The predictions for scenarios 2 and 3 are generally similar and lower than the baseline but with some variation from site to site. Thus the impacts of the reductions in NO X and f-no 2 in scenario 2 result in roughly equivalent ambient NO 2 concentrations to the impact of the reduction in f-no 2 for light duty vehicles assumed for scenario 3. The lowest ambient NO 2 predictions were obtained for scenario 4, which incorporates the reduction in NO X emissions and f-no 2 for heavy duty vehicles from scenario 2 with an additional reduction in f-no 2 for light duty vehicles relative to that assumed in scenario 3 (6% f-no 2 for Euro 6 cars and Euro 5 and 6 vans, compared with 1% assumed in scenario 3 and 55% in the baseline). In terms of achievability of the NO 2 annual mean limit value in 21, scenario 4 has the greatest impact in terms of reducing concentrations by.-.6 μg m -3 relative to the baseline for this year. By 22 the decrease in annual mean concentrations for scenario 4 relative to the baseline is between.9 and 15.8 μg m -3. These differences at some sites will be sufficient to reduce the annual mean concentration below the limit value. However at many sites in Figure 8.2 the concentration is predicted to remain above the limit value even for scenario 4. AEA Energy & Environment 16

118 9 Wider Applicability of Model Results Across the EU 9.1 Introduction The aim of this section is to consider how to apply conclusions drawn from the site-specific model results presented in the previous sections of projected ambient NO 2 concentrations for 21, 215 and 22 to wider areas across the EU. This will involve three stages as follows. Firstly, section 9.2 presents a method for estimating the additional NO 2 annual mean concentration (Delta NO 2 1) likely to occur at a monitoring site where projections have not already taken into account the impact of increasing f-no 2. An approximate relationship has been found between the results of the Netcen Primary NO 2 model for the ten Member States in terms of Delta NO 2 and the baseline annual mean NO 2 concentration. This is presented because it will act as a guide to the amount that projected NO 2 concentrations at monitoring sites where f-no 2 increases have not been accounted for may be under predicting. Section 9.2 also presents Delta NO 2 2, which is the reduction in NO 2 concentrations delivered by the more optimistic scenario 4 relative to the baseline for the Netcen Primary NO 2 model. This is included as it gives an indication of the magnitude of the impact of our most optimistic scenarios on ambient NO 2 concentrations into the future. Section 9.3 then presents further modelling that has been carried out for the UK in order to illustrate how the results for individual sites relate to the situation across this Member State. This involves comparing the NO 2 projections from the Netcen Primary NO 2 model at the case study sites with NO 2 projections generated using the PCM model for these sites, using the same assumptions as used in the Netcen Primary NO 2 model. The PCM model, described in Stedman et al (26), covers all urban major road links in the UK. If there is good agreement at the case study sites then we can look at the NO 2 projections for the other road links not included in the original case study in order to illustrate how the results for individual sites relate to the situation across the whole Member State. Finally, in section 9.4 we test the performance of the Delta NO 2 method by applying this method to the UK PCM modelling results where future changes in f-no 2 have not been included in the modelling. The results of this analysis suggest that this method does provide an approximate way of estimating the impact of changes in the extent of exceedence of the annual mean limit value across a Member State from the site-specific results. 9.2 An Analysis of the Typical Impact of Changes in the Annual Mean NO 2 as a Result of Changes in f-no Introduction The site-specific results from the Netcen Primary NO 2 model have been analysed further with a view to drawing some general conclusions regarding the likely impact of future changes in primary NO 2 emissions on the achievability of the annual mean limit value of 4 μg m -3 at the roadside in Member States. We have done this by calculating the differences (Delta NO 2 1 and 2) between the site-specific predictions for the baseline, sensitivity test and also scenario results. These Deltas have been compared with baseline annual mean NO 2 projections and a relationship has been derived using regression analysis The Impact of Baseline Changes in f-no 2 Figure 9.1a shows a plot of Delta NO 2 1 annual mean NO 2 against baseline annual mean NO 2 predictions for 21. Delta NO 2 1 has been defined as the baseline prediction minus the prediction for AEA Energy & Environment 17

119 the sensitivity analysis in which the emission inventory based f-no 2 for 25 has been applied in all years (see section 7.2 for full details of this sensitivity test). Thus Delta NO 2 1 is the impact of likely future changes in f-no 2 on annual mean NO 2 arising from abatement technologies. This illustrates well that the impact of changes in f-no 2 is likely to be greatest at the sites with the highest concentrations. Two regression slopes have been plotted, one for all of the sites and a second in which the five monitoring sites with the highest NO 2 concentrations (all case study sites in Paris and London Marylebone Road in the UK) have been excluded. The slope of the latter is lower suggesting that the relationship with baseline annual mean is probably not linear. The data are however sufficiently scattered that we do not consider it appropriate to attempt a non-linear fit. Figure 9.1. Delta NO 2 1 annual mean NO 2 21 in 21, 215 and 22 a) Delta NO 2 1 annual mean NO 2 21 (baseline f-no ( 2 sensitivity analysis ywith 25 y f-no 2 ) ) Delta annual mean NO2 (ug m-3) Austria Czech Republic Finland France Germany Greece Italy Netherlands Spain UK all sites lower NO2 sites Linear (all sites) Linear (lower NO2 sites) y =.2922x R 2 = y =.111x R 2 = baseline annual mean NO2 concentration (ug m-3) AEA Energy & Environment 18

120 Figure 9.1 continued b) Delta NO 2 1 annual mean NO (baseline f-no 2 - sensitivity analysis with 25 f-no 2 ) ( y y ) Delta annual mean NO2 (ug m-3) Austria Czech Republic Finland France Germany Greece Italy Netherlands Spain UK all sites lower NO2 sites Linear (all sites) Linear (lower NO2 sites) y =.3891x R 2 = y =.1659x R 2 = baseline annual mean NO2 concentration (ug m-3) c) Delta NO 2 1 annual mean NO 2 22 (baseline f-no 2 - sensitivity analysis with 25 f-no 2 ) ( y y ) Delta annual mean NO2 (ug m-3) Austria Czech Republic Finland France Germany Greece Italy Netherlands Spain UK all sites lower NO2 sites Linear (all sites) Linear (lower NO2 sites) y =.3596x R 2 = y =.1443x R 2 = baseline annual mean NO2 concentration (ug m-3) AEA Energy & Environment 19

121 Table 9.1a shows examples of Delta NO 2 1 annual mean NO 2 calculated using these two regression relationships for different baseline NO 2 concentrations. We consider that the lower slope is appropriate for annual mean concentrates in the range 3 6 μg m -3 and the higher slope from 4 to 1 μg m -3 and above. Thus the range of possible results between 4 and 6 μg m -3 gives some indication of the uncertainties in this attempt to apply the results of our study more widely. This table suggests that the changes in f-no 2 implied by the baseline result in approximately an additional 2 μg m -3 annual mean NO 2 at 4 μg m -3 and μg m -3 at 6 μg m -3 in 21. Figure 9.1a shows a for 215, when Delta NO 2 1 annual mean NO 2 is expected to the larger (a larger impact of changes in primary NO 2 ), that roughly an extra 4 μg m -3 at 4 μg m -3 and μg m -3 at 6 μg m -3 is expected. The impact in 22 (Figure 9.4c) is similar. Table 9.1a Estimates of Delta NO 2 1 annual mean NO 2 (the impact baseline increases in f-no 2 relative to f-no2 in 25) for different baseline concentrations. Derived from the sitespecific results across ten member states (μg m -3 ). Delta NO Delta NO Delta NO Baseline Lower NO annual mean 2 Lower NO All sites 2 Lower NO All sites 2 All sites sites sites sites NO Table 9.1b Estimates of Delta NO 2 1 annual mean NO 2 (the impact baseline increases in f-no 2 relative to f-no2 in 25) for different concentrations in which changes in f-no 2 have not been taken into account. Derived from the site-specific results across ten member states (μg m -3 ). Annual mean NO 2 with constant 25 f-no 2 Delta NO Delta NO Delta NO Lower NO 2 sites All sites Lower NO 2 sites All sites Lower NO 2 sites All sites The results presented in Table 9.1a show estimates of Delta NO 2 1 for different baseline annual mean values (i.e. the NO 2 concentrations quoted for Delta NO 2 1 are already included in the baseline annual mean projection. Delta NO 2 1, in this case, shows how much of the total NO 2 in the baseline projections is caused by f-no 2 changes into the future). Table 9.1b shows estimates of Delta NO 2 1 for values of annual mean NO 2 calculated for the sensitivity analysis in which f-no 2 has been held constant at 25 values for all years. Thus if you have an prediction of annual mean NO 2 that has AEA Energy & Environment 11

122 been derived for 21, 215 or 22 without taking into account the expected changes in f-no 2 after 25 then you can use Table 9.1b as a look up table for Delta NO 2 1. You can then add Delta NO 2 1 to your prediction of annual mean NO 2 to get an estimate that takes into account expected baseline changes in f-no The Impact of Scenario 4 Figure 9.2 and Table 9.2 show the difference between the results for the more optimistic scenario 4 projections and the baseline (Delta NO 2 2). This shows the impact on annual mean NO 2 concentrations of the more optimistic assumptions concerning f-no 2 and a reduction in NO X emissions from HDV at Euro VI. Table 9.2 Estimates of Delta NO 2 2 annual mean NO 2 (the reductions delivered by scenario 4, relative to the baseline) for different baseline concentrations. Derived from the site- specific results across 1 member states (μg m -3 ). Delta NO Delta NO Delta NO Baseline Lower NO annual mean 2 Lower NO All sites 2 Lower NO All sites 2 All sites sites sites sites NO AEA Energy & Environment 111

123 Figure 9.2. Delta NO 2 2 annual mean NO 2 21 in 21, 215 and 22 a) Delta NO 2 2 annual mean NO 2 21 (baseline - scenario 4) ( ).7 Austria Czech Republic.6 Finland France Germany Delta annual mean NO2 (ug m-3) Greece Italy Netherlands Spain UK all sites lower NO2 sites Linear (all sites) Linear (lower NO2 sites) y =.53x R 2 =.28 y =.47x R 2 = baseline annual mean NO2 concentration (ug m-3) b) Delta NO 2 2 annual mean NO (baseline - scenario 4) ( ) 6 Austria Czech Republic Finland 5 France Germany Greece Delta annual mean NO2 (ug m-3) Italy Netherlands Spain UK all sites lower NO2 sites Linear (all sites) Linear (lower NO2 sites) y =.431x R 2 =.2679 y =.447x R 2 = baseline annual mean NO2 concentration (ug m-3) AEA Energy & Environment 112

124 Figure 9.2 continued c) Delta NO 2 2 annual mean NO 2 22 (baseline - scenario 4) ( ) 18 Austria Czech Republic Finland France Germany Greece Italy Netherlands Spain UK all sites lower NO2 sites Linear (all sites) Linear (lower NO2 sites) Delta annual mean NO2 (ug m-3) y =.1233x R 2 =.3254 y =.1982x R 2 = baseline annual mean NO2 concentration (ug m-3) There is little impact of scenario 4 relative to the baseline in 21 so both regression lines (all sites and lower NO 2 sites) are similar, with very low gradients. The impact is somewhat greater in 215 and scenario 4 is delivering a reduction relative to the baseline of about 1 μg m -3 at an annual mean of 4 μg m -3 and about 2 μg m -3 at 6 μg m -3. The impact is considerably greater in 22 and is estimated to result in a reduction of about 4 μg m -3 at 4 μg m -3 and 6 8 μg m -3 at 6 μg m -3 relative to the baseline. Thus the results of our detailed site-specific analyses in ten Member States can be used to derive approximate expected changes in annual mean NO 2 concentrations at roadside locations as a result of changes in primary NO 2 emissions. This impacts (Delta NO 2 1 and 2) are listed in Tables 9.1 and Case Study: An Assessment of the Extent of Exceedences of the Annual Mean NO 2 Limit Value across the UK Introduction The predictions of future ambient NO 2 concentrations presented in this report for scenarios with different values for NO X emissions and f-no 2 have necessarily been focussed on the individual locations of monitoring sites. It is, however, also important to examine the impact of changes in NO X emissions and f-no 2 over a wider range of locations. We have therefore carried out a series of scenario calculations for all of the urban major roads in the UK, as a case study, to illustrate how the results for individual sites relate to the situation across a Member State. AEA Energy & Environment 113

125 9.3.2 PCM Model We have used the PCM model to estimate annual mean NO 2 concentrations at urban major roads in the UK. This model has been described by Stedman et al (26). This is a GIS based model and includes contributions to ambient annual mean NO X concentrations from regional background, an urban increment and a roadside increment. The oxidant partitioning model (Jenkin, 24) is used to estimate annual mean NO 2 concentrations from estimates of NO X, regional oxidant and local oxidant (primary NO 2 ). We have used the version of the model as further developed by Kent et al (in preparation) for a base year of 25 to provide estimates of annual mean NO 2 concentrations in 21, 215 and 22 for a total of 9,937 urban major road links (with a total length of 14,218 km). The model has been used to estimate NO 2 concentrations in the UK for reporting the annual air quality assessment and report on plans and programmes for the 1st Air Quality Daughter Directive. These projections incorporate UK emissions derived from the NAEI (Dore et al, 26) and estimates of f-no 2 derived from the report of the Air Quality Expert Group (AQEG, 26). Table 9.3 shows the values of f-no 2 suggested by AQEG for different vehicle classes and Euro standards. Note that these are broadly similar, but not identical to f-no 2 values that we have adopted in this study. Table 9.3. f-no 2 for different vehicle classes (proportion). Vehicle category M cycles Petrol cars Diesel cars HGV Bus Petrol LGVs Diesel LGVs Pre Euro I Euro I Euro II Euro III.4.4 speed related speed related (.2 to.4) (.2 to.4) Euro IV+.4.4 speed related speed related (.2 to.4) (.2 to.4) Oxidation catalyst - - speed related (.2 to.4) Particle trap Selective catalytic reduction These values of f-no 2 have been combined with the emission inventory for NO X and fleet composition data to estimate values for f-no 2 for each road link in each year. Different fleet characteristics have been assumed for Central London, London and the remaining urban areas in the UK. f-no 2 is assumed to be highest in Central London due to the larger proportion of the traffic flow represented by taxis and by buses fitted with CRTs. Table 9.4 shows the average values of f-no 2 assumed. Table 9.4. f-no 2 for different vehicle classes (percentage). Urban UK London Central London Both the NO X emission inventory and f-no 2 values are different from the baseline for this current study. This is because in the work presented here we had access to a wider selection of emissions testing data than AQEG and we carried out additional analysis of this data. Following informal discussions with major vehicle manufacturers, we adopted a more pessimistic view on, for example, f-no 2 for Euro 4 and Euro 5 diesel cars and LGVs. We considered this appropriate in assessing the possible impact on achievability of the annual mean limit value. We have therefore carried out a range of scenario calculations to explore the sensitivity of exceedence predictions of the annual mean limit values for NO 2 to different NO X emissions and f-no 2 assumptions. AEA Energy & Environment 114

126 9.3.3 Scenarios Modelled The different scenarios investigated using the PCM model are listed in table 9.5. The PCM AQEG scenario includes link-specific estimates of f-no 2 for each year. f-no 2 has been fixed at 25 values in the PCM AQEG 25 f-no 2 scenario for comparison with the sensitivity analysis presented in section 7. The remaining scenarios are directly comparable with those presented in section 8 but in this instance concentrations have been estimated using the PCM model, rather than the Netcen Primary NO 2 model used in section 8. For comparison purposes site-specific results showing our baseline results from the Netcen Primary NO 2 model presented in section 6 for each site are also considered in this analysis. Table 9.5. The scenarios modelled across the UK using the PCM model. f-no 2 Scenario Description Varies on a road link specific basis Constant f- NO 2 across all road links PCM AQEG PCM AQEG 25 f-no 2 PCM baseline PCM scenario 1 PCM scenario 2 PCM scenario 3 PCM scenario 4 NO X projections include the impact of Euro 5 and Euro 6. f-no 2 derived on a link specific basis using estimates of f-no 2 for different vehicle classes and Euro standards provided by AQEG (26) NO X projections include the impact of Euro 5 and Euro 6. f-no 2 set to 25 values for all years NO X projections include the impact of Euro 5 and Euro 6. Baseline f-no 2 taken from this study (same value for all links) NO X projections include the impact of Euro 5 and Euro 6. Scenario 1 f-no 2 (more pessimistic f-no 2 for HDV) taken from this study (same value for all links) NO X projections include the impact of Euro 5, Euro 6 and an estimate for Euro VI. Scenario 2 f-no 2 (more optimistic f-no 2 and NO X control for HDV) taken from this study (same value for all links) NO X projections include the impact of Euro 5 and Euro 6. Scenario 3 f-no 2 (more optimistic f-no 2 LDV at Euro 6) taken from this study (same value for all links) NO X projections include the impact of Euro 5, Euro 6 and an estimate for Euro VI. Scenario 4 f-no 2 (more optimistic f-no 2 and NO X control for HDV and more optimistic f- NO 2 for LDV than scenario 3) taken from this study (same value for all links) Results for Individual Sites Figure 9.3 shows the results for the monitoring sites in London for which modelling using the Netcen Primary NO 2 model has also been carried out. The highest predicted NO 2 concentration in 22 is generally for PCM scenario 1. The lowest predicted concentrations in 22 are generally for PCM scenario 4 (the most optimistic scenario in this study) and the PCM AQEG 25 f-no 2 scenario (in which f-no 2 has been held at the lower 25 values for all the years). The predicted concentrations are generally higher for the PCM baseline than for PCM scenario 1 because the f-no 2 values derived for this study are typically rather higher than those derived from the AQEG recommendations. The exact shape of the curves are dependent on the balance between the roadside increment and background contributions at the different roadside sites. Changes in the f-no 2 assumptions generally have the greatest impact at London Marylebone Road, which is the site with the greatest roadside increment. Site-specific projections calculated from a monitoring base year of 24 using the Netcen Primary NO 2 model for the baseline scenario are also shown on figure 9.3. The agreement between the two models is generally good in 21, 215 and 22 and is very good at London Marylebone Road and London Cromwell Road 2. This suggests that the estimates of NO X concentrations within the PCM model are in reasonable agreement with measurements (the Netcen Primary NO 2 model projections were derived directly from measurement data without the use of an air dispersion model) and that the oxidant-partitioning model responds in a similar way to the Netcen Primary NO 2 model to changes in f-no 2. AEA Energy & Environment 115

127 Figure 9.3. Estimates of annual mean NO 2 at the London case study sites calculated using the PCM model a) Camden Kerbside b) London A3 Roadside 7 6 Camden Kerbside London A3 Roadside 6 5 Annual mean NO2 concentration, ug m pcm_aqeg pcm_aqeg_25f-no2 pcm_baseline pcm_scenario 1 pcm_scenario 2 pcm_scenario 3 pcm_scenario 4 f-no2 varies on a roadlink specific basis f-no2 constant across all road links Annual mean NO2 concentration, ug m Site-specific baseline year year c) London Cromwell Road 2 d) London Marylebone Road 7 9 London Cromwell Road 2 London Marylebone Road 6 8 Annual mean NO2 concentration, ug m Annual mean NO2 concentration, ug m year year e) Southwark Roadside f) Tower Hamlets Roadside 7 7 Southwark Roadside Tower Hamlets Roadside 6 6 Annual mean NO2 concentration, ug m Annual mean NO2 concentration, ug m year year AEA Energy & Environment 116

128 9.3.5 Results for the UK The PCM model has been used to calculate predictions of annual mean NO 2 concentration for the different scenarios for a total of 9,937 urban major road links across the UK. Figure 9.4a shows a summary of the PCM model results for the 2,16 urban major road links in London (which have a total length of 1,89 km) in terms of the percentage of total road length predicted to exceed the annual mean limit value of 4 μg m -3 for the different scenarios. The results follow the same general trend as those presented for individual sites with more exceedences predicted for the baseline and scenario 1 than for the PCM AQEG scenario due to the greater f-no 2 assumed. The percentage of urban major road length in London exceeding in 22 ranges from 23.4% for PCM scenario 1 to 6.7% for PCM scenario 4 (more optimistic than the baseline for LDV and HDV). Thus the extent of exceedence in 22 is found to be highly dependent on both the NO X emissions and f-no 2 Figure 9.4b shows the percentage exceedence for urban major roads in the UK outside London. The estimated percentage exceeding in 22 ranges from 2.5% for PCM scenario 1 to.5% for PCM scenario 4. The extent of exceedences is much lower outside London. Figure 9.2c shows the results for the whole of the UK. Figure 9.4. Percentage of urban major road length with annual mean NO 2 concentrations greater than 4 μg m -3 a) In London only b) Outside London 6% London 14% Outside London 5% 12% 1% 4% percentage 3% 2% pcm_aqeg pcm_aqeg_25f-no2 pcm_baseline f-no2 varies on a roadlink specific basis percentage 8% 6% 4% pcm_scenario 1 1% pcm_scenario 2 pcm_scenario 3 f-no2 constant across all road links 2% pcm_scenario 4 % year c) Across the UK 2% UK % year 16% 12% percentage 8% 4% % year AEA Energy & Environment 117

129 9.3.6 Conclusions The comparison of NO 2 concentration projections from the Netcen Primary NO 2 model and the PCM model at the six sites selected for the London case study showed that they produced broadly comparable results. This means that we can use the PCM model to estimate the likely extent of exceedences for other UK road links not modelled using the Netcen Primary NO 2 model. This is useful for drawing conclusions about how widespread exceedences will be across the UK. The Netcen Primary NO 2 model showed the NO 2 annual mean limit value is likely to be exceeded for all six London sites modelled up to and including 22. The PCM model results suggest that between 6.7 and 23.4% of total urban road length in London will exceed by 22 (all six London case study sites are on road links where the PCM model predicts exceedences will occur). Therefore, in extrapolating the site-specific results that have been the focus of this study to a wider area, it is clear that these sites are on roads that have higher than typical NO 2 concentrations in London or in the UK. This is because the focus of our study has been sites at risk of exceeding the limit value. These results therefore suggest that the exact extent of roads exceeding the limit value will depend on the frequency distribution of roadside NO 2 concentrations in the Member State in question. For many of the Member States selected as case studies, the roadside sites included in our analysis have higher than typical concentrations, so the extent of exceedences away from these sites may be considerably less, as in the UK example. 9.4 Relevance of Typical Impacts to Estimate Extent of Exceedences in the UK and Europe The relationship between changes in concentrations at individual monitoring sites and across all of the roads in a Member State is clearly dependent on the frequency distribution of roadside NO 2 concentrations in that Member State. In this section we compare estimates calculated by applying Delta NO 2 1 and 2, derived above, with the case study results for the UK calculated using the PCM model. We have taken the PCM model results for annual mean NO 2 for the PCM baseline scenario for each of the urban major roads in the UK as the starting point for this analysis. The reason for selecting the PCM model to use here is that it includes all urban road links in the UK and therefore gives a wide spatial coverage for us to test Delta NO 2 1 and 2. We have then used the relationships shown in Figures 9.1 and 9.2 to calculated estimates of Delta NO 2 1 and Delta NO 2 2 for each of these links. Delta NO 2 1 enables us to estimate roadside annual mean NO 2 concentrations from the baseline for the sensitivity test in which f-no 2 is assumed to remain unchanged at 25 levels in all years. This can be compared with the PCM AQEG 25 f-no 2 sensitivity analysis. Delta NO 2 2 is used to estimate concentrations from the baseline analogous to the PCM scenario 4. The estimates calculated using Deltas NO 2 1 and 2 (d_pcm AQEG 25 f-no 2 and d_pcm scenario 4) are shown in Figure 9.5 along with the PCM baseline and PCM AQEG 25 f-no 2 sensitivity analysis and PCM scenario 4 results in terms of the extent of exceedences of an annual mean concentration of 4 μg m -3. We have taken the average of the results of the two regression relationships for Delta NO 2 1 and 2 for baseline annual mean NO 2 concentrations in the range from 4 to 6 μg m -3. Thus we have derived estimates of the extent of exceedence both by applying the typical Delta NO 2 1 and 2 values from the site-specific modelling and by carrying out full PCM calculations for each road link. The expressions used to calculate the estimates are listed below: d_pcm AQEG 25 f-no 2 = pcm_baseline Delta NO 2 1 d_pcm scenario 4 = pcm_baseline Delta NO 2 2 The agreement between the two methods is reasonably good for the 25 f-no 2 sensitivity with the method of applying Delta NO 2 1 estimating slightly fewer exceedences. Thus Delta NO 2 1 is providing a reasonable description of the impact of changes in f-no 2. AEA Energy & Environment 118

130 The impact of scenario 4 relative to the baseline seems to be somewhat underestimated by applying Delta NO 2 2 relative to the PCM calculations but the general trends are reasonably well represented. It is clear from Figures 9.3 and 9.4 that there is scatter in the relationship between the Deltas and annual mean NO 2 but that the impact is clearly greatest at the roads with the highest baseline concentrations. Overall the agreement between the two methods (applying Deltas NO 2 1 and 2 and the full PCM model calculations) is however, reasonably good. This suggests that this method of generalising the results of the site-specific modelling using Delta NO 2 1 and 2 does provide an approximate way of estimating the impact of changes in exceedence of the annual mean limit value across a member state from the site-specific results in this study. Figure 9.5. Percentage of urban major road length with annual mean NO 2 concentrations greater than 4 ug m -3 a) In London only b) Outside London 6% 14% London Outside London 5% 12% 1% 4% percentage 3% percentage 8% 6% 2% 1% pcm_aqeg_25f-no2 pcm_baseline pcm_scenario 4 d_pcm_aqeg_25f-no2 d_pcm_scenario4 4% 2% % year % year AEA Energy & Environment 119

131 1 Concluding remarks The impact of primary NO 2 emissions on ambient NO 2 air quality is expected to be most pronounced at roadside locations with the highest NO X concentrations. A combination of emission inventory calculations and projections of ambient air quality suggest that the impact will be greatest between 25 and 215. Changes in vehicle exhaust after treatment technology, particularly selective catalytic reduction are expected to result in a decrease in the emissions of primary NO 2 and thus the impact on roadside air quality by 22. The actual magnitude of primary NO 2 emissions and thus NO 2 concentrations and the extent of exceedence of the annual mean limit value in 22 will be dependent on the exhaust limit values set and the technology adopted, as illustrated by our scenario calculation results. It is likely that exceedence of the limit values in 21 will be considerably more widespread across the EU than originally envisaged as a result of changes in f-no 2 between 2 and 21. AEA Energy & Environment 12

132 11 Acknowledgements This work was funded by the European Commission. Contract number 751/26/438919/MAR/C3 AEA Energy & Environment 121

133 12 References Abbott, J. (25) Primary nitrogen dioxide emissions from road traffic: analysis of monitoring data. AEA Technology, National Environmental Technology Centre. Report AEAT Airbase (26) Airparif (26) Air Quality Expert Group (AQEG 26) Primary NO 2 in the United Kingdom (in consultation) AQS (26) The Air Quality Strategy for England, Scotland, Wales and Northern Ireland. A consultation document on options for further improvements in air quality. Volume 2: Technical annex and regulatory impact assessment. April 26. Carslaw, D.C. and Beevers, S.D. (25) Estimations of road vehicle primary NO 2 exhaust emission fractions using monitoring data in London. Atmospheric Environment 39 ( ) Central Data Repository (CDR 26) Collier C G, Norris JOW and Murrells T P (25) Analysis of Measured Emission Factors for Euro III Cars and their Incorporation into the National Atmospheric Emissions Inventory, report to DfT, netcen/4868/euro III EF report final_nov5, AEAT/ENV/R/283, Council Directive 7/22/EC of 2/3/197. On the approximation of the laws of the Member States relating to measures to be taken against air pollution by emissions from motor vehicles. Council Directive 88/77/EEC. On the approximation of the laws of Member States relating to the measures to be taken against the emissions of gaseous and particulate pollutants from diesel engines for use in vehicles. Council Directive 91/441/EC of the European Parliament and of the Council relating to measures to be taken against air pollution by emissions from motor vehicles. Official Journal of the European Communities OJ L242, 1991 Council Directive 94/12/EC of 23/3/1994. Relating to measures to be taken against air pollution by emissions from motor vehicles and amending Directive 7/22/EC Council Directive 96/62/EC, of 27 September On ambient air quality assessment and management (The Framework Directive). From the Official Journal of the European Communities, , En Series, L296/55. Council Directive 98/69/EC of 13/1/1998. Relating to measures to be taken against air pollution by emissions from motor vehicles and amending Council Directive 7/22/EEC Council Directive 1999/3/EC, of 22 April 1999 Relating to limit values for sulphur dioxide, nitrogen dioxide and oxides of nitrogen, particulate matter and lead in ambient air (The First Daughter Directive). From the Official Journal of the European Communities, , En Series, L163/41. Council Directive 99/96/EC On the approximation of the laws of Member States relating to the measures to be taken against the emissions of gaseous and particulate pollutants from compression ignition engines for use in vehicles, and the emission of gaseous pollutants from positive ignition engines fuelled with natural gas or liquefied petroleum gas for use in vehicles and amending Council Directive 88/77/EEC. Departament de Medi Ambient i Habitatge (Generalitat de Catalunya) (26) Personal communication AEA Energy & Environment 122

134 Derwent, R.G., Simmonds, P.G., O Doherty, S. et al., (25) External influences on Europe s air quality: baseline methane, carbon monoxide and ozone from 1999 to 23 at Mace Head, Ireland. Atmospheric Environment, 4, Dore, C. J., Watterson, J. D., Goodwin, J. W. L. et al., (26). UK Emissions of Air Pollutants 197 to 24. National Atmospheric Emissions Inventory, AEA Technology, AEA Energy & Environment GAINS (26) from IIASA website GLA (26) London Atmospheric Emissions Inventory 23 Jenkin, M. E. (24). Analysis of sources and partitioning of oxidant in the UK-Part 1: the NOxdependence of annual mean concentrations of nitrogen dioxide and ozone. Atmospheric Environment Kessler, C., Niederau, A. and Scholz, W. (26) Estimation of NO 2 /NO X relations of traffic emissions in Baden-Wurttemberg from 1995 to nd conf. Environment & Transport, incl. 15 th conf Transport and Air Pollution. Reims, France, June 26. Proceedings no 17, vol. 2, Inrets ed., Arceuil, France, 26, p Kent, A.J., Grice, S. E., Stedman, J. R., Bush, Vincent, K. J., Abbott, J. and Derwent, R. G, Hobson, M. (in preparation) UK air quality modelling for annual reporting 25 on ambient air quality assessment under Council Directives 96/62/EC, 1999/3/EC and 2/69/EC. AEA Technology, AEA Energy & Environment Report AEAT/ENV/R/2278 Issue 1 Landesanstalt für Umwelt, Messungen und Naturschutz Baden-Württemberg (LUBW 26) Lambrecht, U., (26) Findings in Germany NO X : Development of emissions and air quality. Causes for unexpected values of ambient concentrations of NO 2. EU level workshop on the impact of direct emissions of NO2 from road vehicles on NO 2 concentrations. Brussels, September 19th, 26 cles/eu-workshop_ /_en_1._&a=d Latham S, Kollamthodi S, Boulter P, Nelson PM and Hickmann AJ. Assessment of primary NO 2 emissions, hydrocarbon speciation and particulate sizing on a range of road vehicles, TRL ref nos PR/SE/353/21 Stedman, J. R., Bush, T. J., Grice, S. E., Kent, A. J., Vincent, K. J., Abbott, J. and Derwent, R. G. (26). UK air quality modelling for annual reporting 24 on ambient air quality assessment under Council Directives 96/62/EC, 1999/3/EC and 2/69/EC. AEA Technology, AEA Energy & Environment. Report AEAT/ENV/R/252 Stones P, Sandbach E, Norris JOW (26) The effects of emerging vehicle technologies on certain vehicle emissions not currently regulated, DfT CFV Project Reference Number S426/T4, Contractor - Millbrook Proving Ground Ltd, Report No. MBK 5/832 TfL (26). Reported in AQEG Trends in Primary Nitrogen Dioxide in the UK, Draft report for comment, August 26 (TfL exhaust emission data. Personal communication between Anna Rickard (TfL) and Ian McCrae (TRL), 5/4/6). The AQEG consultation draft can be found at TREMOVE 2.44, which has been developed by Transport & Mobility Leuven and the K.U.Leuven in a service contract for the European Commission, DG Environment. Traffic Management and Air Quality Programme (TRAMAQ) UK Department for Transport, -, UG216 AEA Energy & Environment 123

135 Werner Scholz pers. Comm. 27 UK Air Quality Archive (26) AEA Energy & Environment 124

136 AEAT/ENV/R/2278 Issue 1 AEA Energy & Environment