Evolution of NO x emissions in Europe with focus on road transport control measures

Abstract. European emission trends of nitrogen oxides since 1880 and up to present are presented here and are linked to the evolution of road transport emissions. Road transport has been the dominating source of NO x emissions since 1970, and contributes with 40% to the total emissions in 2005. Five trend regimes have been identified between 1880 and 2005. The first regime (1880–1950) is determined by a slow increase in fuel consumption all over Europe. The second regime (1950–1980) is characterized by a continued steep upward trend in liquid fuel use and by the introduction of the first regulations on road traffic emissions. Reduction in fuel consumption determines the emission trends in the third regime (1980–1990) that is also characterized by important differences between Eastern and Western Europe. Emissions from road traffic continue to grow in Western Europe in this period, and it is argued here that the reason for this continued NO x emission increase is related to early inefficient regulations for NO x in the transport sector. The fourth regime (1990–2000) involves a turning point for road traffic emissions, with a general decrease of emissions in Europe during that decade. It is in this period that we can identify the first emission reductions due to technological abatement in Western Europe. In the fifth regime (2000–2005), the economic recovery in Eastern Europe imposes increased emission from road traffic in this area. Western European emissions are on the other hand decoupled from the fuel consumption, and continue to decrease. The implementation of strict measures to control NO x emissions is demonstrated here to be a main reason for the continued Western European emission reductions. The results indicate that even though the effectiveness of European standards is hampered by a slow vehicle turnover, loopholes in the type-approval testing, and an increase in diesel consumption, the effect of such technical abatement measures is traceable in the evolution of European road traffic emissions over the last 15 years.

evolved since the 1880s and investigates to what extent the decrease in emissions after 1990 can be linked to policy regulations. Our analysis links NO x emission trends in Europe to the evolution of fuel consumption as well as to the changes in vehicle technology. It further distinguishes between the Eastern and Western European regions, where differences in the level of penetration of policy measures have an impact 5 on the evolution of the emissions. Although the analysis covers a 125 years time span, the main focus is on the last 15 years, when European NO x emissions have begun to decrease. The methodology developed is presented in Sect. 2 which also documents the data sources used in the analysis. Data quality is discussed in Sect. 3. Results on European trends in NO x emissions are presented in Sect. 4, and the discussion 10 on the effectiveness of policy measures is given in Sect. 5. Finally, conclusions are summarized in Sect. 6. Emissions from international shipping at European waters are analyzed in a forthcoming paper (Jonson et al., in preparation) and are thus not included here. With respect to the terminology adopted, NO x emission figures correspond to NO 2 -equivalents, except in cases where primary NO 2 emissions are explicitly 15 discussed.

Methodology and data used
A European NO x emission inventory that spans over 125 years has been compiled here to provide further insight in the evolution of European road traffic emissions. The inventory relies on available information on 1) activity data, 2) emission factors, 3) 20 abatement level, and 4) the level of policy penetration. Such information is largely variable from period to period and for the different European countries and areas, and determines to a large extent the accuracy of the final results. Concise information on the information sources and an evaluation of the uncertainty associated with each source is also included in the following. Introduction We have estimated anthropogenic fossil fuel and residential wood combustion emission by European country and sector every 5 years since 1880. The underlying activity data corresponds to the European historical country borders in the time span considered, and we distinguish between three different periods: 1880-1915, 1920-1945 and 1950-5 1985. The emission estimation methodology differs only for the periods 1880-1945 and 1950-1985. An advantage of this study with respect to previous global estimates (e.g. van Aardenne et al., 2001;Schultz et al., 2007) is that we have applied emission factors which vary with time and by country. Fuel consumption in the period 1880-1945 is calculated from energy and indus-10 try statistics collected by Mitchell (1981) and supplemented with information from the World Power Conferences (1948). According to international consumption and production statistics, fuel consumption in this period is dominated by solid fuel use. Detailed activity data and reliable emission factors are not available for this period, thus we derive the emissions by scaling 1950 emission sectors per country backwards in time, 15 based on the solid and liquid fuel consumption. We underline that one main implication is that possible important changes in the average emission factor for coal combustion during this period are not considered in our emission estimates. In addition to the emissions from fossil fuel combustion, we also estimated biomass emissions in the residential sector. The coverage of wood consumption per country 20 is rather incomplete in international statistics, particularly before 1950. In addition, we considered that underreporting of wood consumption by countries could be an issue. Therefore we made two different estimates, one based on an extrapolation of OECD activity data (OECD, 2004), and a second one assuming that each person living in rural areas consumes 1 m 3 wood per year, with a conversion factor of 440 kg wood/m 3 (Lun-Introduction  (Sternhufvud et al., 2004). The results from this evaluation are addressed in Sect. 4. With regard to the period 1950-1985, we distinguished between the originally OECD and non-OECD countries, based on the availability of activity data. A detailed breakdown of activity data is published in the OECD Energy Statistics (OECD, 1966(OECD, , 2004, and we estimated NO x emissions from about thirty different sub-sectors aggregated 5 per SNAP sector (Energy Production from Power-plants-S1, Residential-S2, Industry-S3 and S4, Road transport-S7 and Off-road machinery-S8). For the non-OECD countries we used the production figures of electric power in thermal power plants from the UN Energy Statistics (UNECE, 1976(UNECE, , 1980(UNECE, and 1981, to deduce the amounts of lignite and other fuels used in electric power plants. For most countries it has been assumed 10 that mainly lignite was used in electric power plants. However, hard coal needs also to be included in energy budget of Poland and Hungary, while the Former Soviet Union (USSR) and Romania have been also using coal and natural gas, according to the statistics.
Coke production figures were used to deduce the amount of coal used for coke. The 15 remaining coal was distributed between the industrial and the residential sector. For the non-OECD countries, gasoline was assumed to be consumed only in cars, or in other internal combustion engines. It was also assumed that gas oils must have been too expensive to be used except in internal combustion engines of cars, trucks, off-road equipment and machinery, and in agriculture. In general, these uses are assumed to 20 account for 80% of the gas oil consumption. The emission factors used for this period are shown in Table 1 and are broadly based on the work by Pacyna et al. (1991), reviewing a large selection of country specific emission factors from national and international programmes, with a special attention on Eastern Europe. The emission factors from Pacyna (1991)  Interactive Discussion itate crude conversion of the emission factors used in this study to other units, we have also indicated average calorific values in brackets in Table 1.
One main point is that the emission factors for motor vehicles increase with time during the period 1950 to 1985. This is to reflect the fact that the development of new gasoline engines over this period led to less fuel-enriched mixtures, lower scaveng-5 ing losses, and higher compression rates to improve fuel efficiency and to control CO and HC emissions that were the focus at the time (UNECE, 1958). As a side-effect of improved combustion, engines led to increased (thermal) NO x emissions to the atmosphere. According to Samaras and Zierock (1996), emission factors of 20.4 g/kg were applicable for gasoline cars without emission controls, i.e. vehicles produced before 10 1970 (Pre ECE R-15) but for vehicles with non-catalyst control (i.e. improved combustion) the emission factors increased to 36.7 g/kg (Table 1 footnote 3).
An increase in emission factors also occurred for diesel engines during this period, according to the US-AP42 (US EPA, 1991). The low emission factors for diesel engines are typical of engines with indirect injection. This design is not favoured for 15 modern, large trucks, which have direct-injection engines and higher compression ratios. The effective compression ratio may be further increased by turbo-charging, which further promotes the formation of NO x . Emissions from for heavy duty vehicles (HDV) were not regulated until 1988 with the introduction of the ECE 49 Regulation. The lower limit of the emission factor range for HDVs in Table 1 (30 g/kg) is comparable   20 to Conventional HDVs included in COPERT, when these are converted per fuel mass used, (http://lat.eng.auth.gr/copert) and, further, to the uncontrolled NO x emission factors in the GAINS database (http://www.iiasa.ac.at/web-apps/apd/gains) developed at IIASA. Measurement studies more often concern US and more recent vehicle fleets (e.g. Kirchstetter et al., 1999;Yanowitz et al., 2000;Kristensson et al., 2003;Schmid et 25 al., 2000;Kohler et al., 2004), but Ekström et al. (2004) report on-road optical remote sensing measurements in Sweden per vehicle technology class which support the upper limits of the emission factor ranges both for gasoline passenger cars and heavy duty vehicles in Table 1 The emission factors we have assigned to HDVs are lower in Eastern European (24-40 g/kg) than Western European countries (30-50 g/kg), to reflect the differences in vehicle technologies following the implementation of the ECE-R15 regulations in Western Europe. In addition, some Eastern European countries used to have a high proportion of 2-stroke engines which resulted in even lower average NO x emission 5 factors. In the extreme case of Former East Germany, more than 50% of the vehicles used to have 2-stroke engines, and the resulting emission factor for gasoline cars is consequently estimated at a much lower value (6-10 g/kg) than for the other countries included in this study (20-30 g/kg). While these simple considerations do not provide detailed and accurate inventories for each country, they may still give useful estimates 10 of regional and temporal trends.
For stationary sources we do not include any variation of emission factors with time. We assign country specific emission factors for brown coal in thermoelectric power plants and in the industry according to fuel quality and combustion technology in the respective countries (UNECE, 1981;McInnes, 1996). This implies that emission factors 15 for power plants are generally higher in Eastern Europe compared to Western Europe. The highest emission factor is assigned to Former Czechoslovakia (8 g/kg), followed by Albania, Bulgaria, Former USSR and Yugoslavia (7 g/kg), Poland (6 g/kg), Former East Germany, Hungary, Romania, Austria, Denmark and France (5 g/kg), Spain (4 g/kg), former West Germany Italy, Portugal and Turkey (3 g/kg) and Greece (2 g/kg). Due to 20 lack of information about differences between countries in the industry sector, we have applied a uniform emission factor of 3 g/kg for the industry sector in all Eastern European countries. Emission factors for Western European industries are about 1 g/kg lower than those listed above for power plants.
It is worth noting that we do not include the gas associated with the production of 25 coke from coal, hence available for combustion (e.g. coke oven gas or blast furnace gas from the iron and steel industry) in separate sectors. It is instead included in the emission factor for coke production itself, and this is why we apply an emission factor for coke production orders of magnitude larger than Pacyna et al. (1991) in the residential sector is assumed to occur mainly in small domestic boilers, and the emission factor chosen for combustion in oil refineries is taken from Takacs et al. (2004). Emissions from international shipping and aviation are not included in this study. Further we do not include emissions not directly related to fuel consumption like nitric acid 5 and fertilizer production. According to Pacyna et al. (1991), these are minor sources (0.5% contribution around 1980), as is agricultural burning of straw and stubble (less than 1%). Emissions from waste, which, according to data reported to the LRTAP Convention, contribute less than 1% to the national total in the 1980s is not included. No attempt has been made to include NO from soils although some authors (e.g. Stohl et 10 al., 1996) argue that the emissions are mainly from arable land and should be therefore considered as anthropogenic.

1980-2005: EMEP NO x inventory
For data on NO x emissions after 1980 this study relies mainly on data from the EMEP (Cooperative Programme for Monitoring and Evaluation of the Long-range 15 Transmission of Air Pollutants in Europe) programme. The EMEP inventory consists as far as possible of official data reported annually by 51 Parties to the Convention on Long-range Transboundary Air Pollution (LRTAP). The emission data is compiled at national level in accordance with the UNECE Emission Reporting Guidelines (UNECE, 2003) and the EMEP/CORINAIR Guidebook (http://reports.eea.europa.eu/ 20 EMEPCORINAIR4/en). The national emission estimates are accompanied by an Informative Inventory Report (IIR) documenting the uncertainties in the data used and possible deviations from the recommended methodologies in the Guidebook. These emissions are annually reviewed and evaluated, to check for errors and identify areas where improvements may be necessary (e.g. Vestreng et al., 2007a) Interactive Discussion motorcycles. Emission data available at this level of detail is mainly restricted to the period from 1990 onwards. Several countries report data on a more aggregated level up to 2000, i.e. total Road Transport (SNAP 7). The lack of detailed emission data in the 1980s and partly in the 1990s, has limited parts of our analysis to the aggregated SNAP 7 level. In addition to emission data, national reporting includes activity data for the historical years 1990, 1995, 2000 and 2005. These official activity data reported by countries have been used for the study of emission trends in the period 1990-2005. In the absence of reported data, our analysis for the period 1990-2005 relies on trends in fuel consumption and implied emission factors from the GAINS database. Historic fuel consumption data in GAINS are extracted from national and international energy 10 statistics. For the period 1980 to 1990, we include fuel consumption data from the sources outlined above for the EURONOX inventory. Table 2 presents national NO x emission trends and gives an overview of the completeness of official emissions in the EMEP inventory between 1980 and 2005. The relative share of emissions from road transport (in brackets) is also listed. The emis- 15 sions are based on officially reported data, to the extent possible. Countries which passed the EMEP review are highlighted with grey background in Table 2; a total of nineteen countries. The table identifies also a second group of countries for which reported data had to be completed by interpolation and extrapolation in order to achieve full emission trends for the period. These twelve countries are marked in bold ital-20 ics. For the remaining countries, emissions were derived from other sources. The main source for non-offical emission estimates in the EMEP inventory is data from the RAINS/GAINS model (http://www.iiasa.ac.at/web-apps/apd/gains) developed at IIASA. These emissions are not completely independent from those officially reported in that IIASA, through bi-lateral consultations, may include data provided by the countries 25 themselves. The RAINS/GAINS model is now capable of reproducing national emissions for NO x for almost all Parties with an uncertainty margin of less than 5% (UNECE, 2007). Because there are a few countries for which neither official nor RAINS/GAINS data are available, EDGAR emission data (http://www.mnp.nl/edgar) are also included. Introduction The EDGAR inventory is a global inventory but it does not have the same level of detailed vehicle classification as the RAINS/GAINS model. There are nine countries for which RAINS data have been used and these are marked with a star in Table 2. The emission estimates for the remaining four countries rely on EDGAR emission data. The coverage of reported emissions is about 40% in the 1980s, increasing to nearly 60% 5 after 1990. The level of confidence is considered to be higher for the reported and reviewed emission data, due to country specific insight and the detailed input to the calculations.

Data quality
As indicated from the discussion above, the level of accuracy of the data used all 10 through the NO x inventory from 1880 to 2005 increases as we approach recent times. In this section we document the uncertainties in the EMEP and EURONOX inventories, and justify the merging of these two inventories.

Uncertainties in the EMEP inventory
There are recognised uncertainties in the selection of emission factors and even though 15 national statistics of activity data as compiled from e.g. data reported by individual facilities, registration offices and different surveys are in most cases reliable, there is also an element of uncertainty in this basic input to the national emission calculations. Discrepancies between actual and apparent national emission estimates are also introduced when emission data are reported, in line with the reporting Guidelines, based 20 on fuel sold rather on fuel used. This is because the amount of fuel sold in a country may be strongly influenced by "fuel tourism". This is a term used for retail purchase of fuel in one country for consumption abroad, mainly due to fuel price differences.
Although emissions associated to the combustion of this fuel do not occur in the country where the fuel was sold, the UNECE reporting Guidelines require Parties to report Introduction Interactive Discussion these emissions as national emissions. The effect of this fuel tourism is shown to have opposite and equally large effect for countries with high "green taxes" as discussed in Sect. 5 in the case of Germany. The implication is that while the European emission trend for NO x may be correctly reflected, the national (road transport) emission trends for several European countries may be affected by the tax and transit levels. 5 The uncertainty level for national inventories included in the EMEP inventory is based on information given in the Informative Inventory Reports (IIR) accompanying the official submissions emissions under the LRTAP Convention. Based on a review of this information from a limited number of countries, the uncertainty in national emissions is considered to be between 8% and 23% for Western Europe and around 25% for 10 Eastern Europe. The EMEP inventory contains in addition emission estimates from the RAINS/GAINS model, and according to Schöpp et al. (2005), the uncertainty in these emissions are comparable to those reported by the countries. Uncertainty estimates for individual sectors are not reported by the Parties, but Schöpp et al. (2005) indicate that the sector uncertainty is higher, and might be nearly three times larger 15 for emissions from gasoline passenger cars and diesel heavy duty trucks. Kuhlwein and Friedrich (2000) estimate the statistical error in transport NO x emissions in West Germany to be 16-22%, comparable to the results for United Kingdom estimated by Schöpp et al. (2005). The above uncertainty ranges are applicable from 1990 onwards (Vestreng et al., 2006). Quantitative uncertainty estimates for the 1980s are not avail- 20 able, but they are likely to be larger, due to the lower coverage of reported emissions and absence of published non-official estimates. In addition, recalculation of emission data by many Parties are only performed from 1990 onwards, hence the accuracy in the 1980s emissions may not benefit from methodological improvements in emission estimation. 25 A complementary way to assess the validity of emission data is to combine model and observation data. The general downward trend in EMEP emission data from 1990 onwards have been confirmed by a recent model study by Jonson et al. (2006). The study concludes that even though the EMEP model tends to overpredict winter concen- Interactive Discussion trations and underpredict summer concentrations compared to measurements, NO 2 levels and seasonal patterns are well captured. Further, Fagerli and Aas (2008) show that the reduction in EMEP NO x emissions between 1990 and 2003 are comparable to the downward trend in measurements of nitrate in precipitation. A trend study by Konovalov et al. (2008) applying inversion techniques with GOME 5 and SCIAMACY measurements between 1996 and 2004, broadly confirms that the NO x emission trends in Europe have been decreasing, and further indicates that the quality of the EMEP inventory has increased over the last few years. Our evaluation of regional differences in inventory uncertainties is in agreement with the above study, where particularly large differences between the EMEP and satellite data are found 10 in Balkan countries, Georgia, Russia and Turkey. It is however important to consider that there are limitations both in the model and in the observations when such topdown methodologies are applied to derive emission trends, and we suspect that the lack of reduction in emissions seen for Italy in Konovalov et al. (2008) could also be due to limitations in the vertical extent of the chemical transport model applied for the 15 calculations.

Uncertainties in the EURONOX inventory
The quality of the EURONOX inventory is much influenced by availability and accuracy of the activity data. There are some gaps in the statistical activity data, but in general the quality of the emissions is better from 1950 onwards than before 1950, when less 20 detailed data was available and emissions had to be extrapolated backwards in time. Also, the uncertainty is higher for Eastern than for Western European countries, due to the lack of activity data in specific consumption sectors. There are also uncertainties linked to the selection of emission factors for different sectors and the choice of method, e.g. more generalized emission factors. Emission measurements from stationary sources have only become available in the last 30 years, while measurements of mobile sources came even later. The quality of the emission estimates is hence expected to increase as we approach recent times. The comparability between the EURONOX and EMEP inventories has been assessed for the two common years, 1980 and 1985. Table 3 shows differences in national total and road transport emissions per country in both inventories. The agreement on the national total levels is generally good, with an underestimation of the European total in EURONOX relative to EMEP by 8% and 6% in 1980 and 1985, re-5 spectively. The better agreement in 1985 is probably because the emission factors applied are more representative for 1985 than for 1980. The general overestimation indicates that the emission factors do not reflect any of the emission controls already effective in the 1980s. Particularly large differences between reported and estimated emissions occur for Czechoslovakia, where we apparently underestimated the emis-10 sions by a factor 2. This occurs in spite of our use of emission factors which take into account both the very high energy content of Bohemian lignite and the combustion technology in the Czechoslovakian power plants. If the reported Czechoslovakian NO 2 emissions are correct, the NO 2 /capita ratio in 1980 would have been 74 kg/capita, i.e. far outside the NO 2 /capita range of 20-45 kg elsewhere in Europe, and suggests 15 that the sum of the reported Czech and Slovakian NO 2 emissions are too high. Kopáek and Veslý (2005) have estimated the Czechoslovakian emission to about 710 Gg, thus is more of support of our calculations. The differences in emission estimates for Sweden are also quite large (factor 1.5). The use of gas oils in off-road machinery could explain the underestimation relative to the official emission in this case. Such large 20 differences will be discussed with the national emission experts during the forthcoming reviews.
The comparability is lower when sector data are considered, but the differences are still mostly within the uncertainty range indicated by Schöpp et al. (2005). In contrast to the national total emissions, road transport is generally increasingly overestimated by ACPD 8,2008 European road transport control measures Interactive Discussion much more detailed information about the conditions in each country would have been desirable when developing the EURONOX inventory, the agreement with the EMEP data is considered sufficient to merge the two inventories in 1980, by scaling the EU-RONOX inventory to the relevant EMEP sectors. In order to account for the sources not included in the EUONOX inventory, we scaled the residential sector also together 5 with the EMEP agricultural and waste emissions.

Comparison with other estimates
The combined EURONOX and EMEP inventory is compared to independent inventories both at national (anthropogenic) total and at road-transport levels. The preferred option would have been to make the comparison by country, but the other inventories 10 were not available in such detail. Schöpp et al. (2003) has compiled a NO x inventory based on a study by Dignon and Hameed (1989) between 1880 and 1960. The Dignon and Hameed (1989) inventory is merged with estimates from an old version of the RAINS model from 1960 onwards. The European NO x trend presented in Schöpp et al. (2003) differs considerably from our work in that emissions are consistently higher 15 in Schöpp et al. (2003) over the whole . The difference between the inventories is particularly large in the 1950s, and amount to nearly 40% in 1960 at the European level. Dignon and Hameed (1989) derive emissions by regression analysis from total fuel consumption. It is likely that our refined approach, with application of representative emission factors in distinct fuel consumption sectors, is the main reason 20 for the large discrepancy between these two inventories. A global inventory published by van Aardennne et al. (2001) is available for the years 1890 to 1990 in ten-year intervals. The road-transport emissions are not separately documented but are included in the fossil fuel combustion sector. National and roadtransport data per decade between 1960 and 2000 were made available to us on a 25 regional level for the more recent RETRO inventory (Schultz et al., 2007). EDGAR data (http://www.mnp.nl/edgar) are available per country and sector in five-year intervals between 1990 and 2000. Comparison with these three inventories has been 10712 Introduction Interactive Discussion made on the regional "OECD Europe" and "East Europe" level, as defined by EDGAR (http://www.mnp.nl/edgar). The comparison is made from 1920 onwards as such regional comparisons are hampered by differences in country borders, particularly before 1920.
The OECD emissions presented in Fig. 1 represent between 97% (1920) and 55% Since detailed activity data is available for this period, the difference is apparently due inventory compared to EMEP can only be partly explained by the incompleteness in the RETRO inventory with respect to national navigation, railway, waste treatment and disposal, and cement manufacturing. The difference in trend between the RETRO and the other inventories between 1980 and 2000 seems to be due to application of more efficient abatement in stationary sources, as the trend in road transport compares fairly 20 well both by our work and the EDGAR inventory. Based on these considerations, we conclude that our work is closer to the van Aardenne et al. (2001) and EDGAR estimates than the RETRO emissions at the total OECD level, even though the RETRO inventory for the road transport sector are similar to ours. The much larger relative differences in emission level (more than 100% in some 25 years) and trends for both total and road transport emissions in "East Europe" confirm that the uncertainties are larger in this area (Fig. 2). The comparison indicates an overestimation of emission totals in our study between 1980 and 1990 comparable to the anticipated too high reported emissions from the Czech Republic and Slovakia previ- Interactive Discussion ously discussed. Road transport emissions increase in the EDGAR inventory between 1995 and 2000, contrasting the EMEP emissions. This increase is reflected in the totals, and results in an overestimation of the EDGAR emissions in year 2000 compared both to our study and to the RETRO emissions. The increase in transport emissions in EDGAR follows the trend in fuel consumption in this area, and does not seem to take 5 into account that emissions have decreased in line with the implementation of Euro standards in countries like Poland and the Czech Republic as discussed below. The RETRO road transport trend is much weaker than in our work, indicating that emission factors vary less with time.

Results on European emission trends 1880-2005
10 4.1 European total trends Figure 3 shows the trends in solid and liquid fuel consumption from 1880 to 2005 as compiled for this study. The total fuel consumption increased by more than a factor of ten over a period of a hundred years . Before 1950, solid fuel was the main energy carrier in Europe, and the consumption increased steadily from 1880 onwards; following the oil crises (e.g. Glover and Behrens, 2006) and is also due to decreased  Fig. 4, where also trends per sector are included. The sectors included in this analysis broadly follow the SNAP categorization, i.e. road transport, energy production from power plants, industry, off-road and residential. NO x emissions from residential wood combustion (not shown separately in Fig. 4) only contribute in appreciable amounts to the total NO x emissions around the turn of the 19th century when 10 the contribution is estimated to be around 14%. From 1900 and onwards emissions from biomass have been minor (1-9%). Our results indicate that the residential wood consumption statistics are reasonable compared to our "rule of thumb" estimate. National shipping and domestic aviation is included in the off-road sector. Agriculture and waste, being in general minor NO x sources, have been merged with the residential Interactive Discussion decrease in residual oil for heating plausible, we suspect that the detailed statistics we have used on diesel consumption prior to 1970 might be defective.
In the third regime, 1980-1990, the share of NO x emissions from road transport is large (about 40%), and has remained relatively constant at the European level for the last 25 years (Table 2 and Fig. 4). Total NO x emissions peaked in 1990, partly due 5 to continued increase in road transportation activity up to this point in time, and partly to the fact that emissions from stationary sources remained relatively stable between 1980 and 1990. The fourth regime, 1990-2000, is characterized by a steep decline in NO x emissions. The highest share of road transport to the total emissions (42%) is found around year 2000, and does not coincide with the peak in total NO x emissions. This is due to the slower reduction rate of road-transport emissions (22%) relative to emissions from power plants (42%) and the industry (33%), between 1990 and 2000. As a result of the combined reductions, the total NO x emissions monotonically decreased by 32% between 1990 and 2005. The largest reductions took place in the first half of the 1990s. The reasons of this decline are different in different parts of Eu-15 rope ad will be explained in the next section. Finally, in the fifth emission trend regime, 2000-2005, the downward emission trend has flattened out.

Trend differences between European countries in the last twenty-five years
We focus our analysis on the last three emission trend regimes, i.e. 1980-2005 for two main reasons. First, large changes in the emission trends can at least partly be asso-20 ciated to the technological development and policy regulations in this period. Second, the emission data uncertainty should be lower than in the period before 1980, as indicated in Sect. 3, and this may lead to more solid conclusions. The analysis particularly addresses road-transport, which is the most significant sector and a number of policy regulations have been developed to abate NO x emissions from vehicles. The effectiveness of these regulations in Eastern and Western European countries is separately assessed.
The Interactive Discussion and road transport emissions (percentage contribution in brackets) are highlighted in Table 2. The largest contributors are the Russian Federation, United Kingdom and Germany. The total NO x emissions in Europe increased by 5% from 1980 to 1990 due to increased emissions in most countries but a few notable exceptions like Germany and France where emissions from power plants and the industry were reduced.
5 NO x emissions decreased in most countries between 1990 and 2005, but there are substantial differences in the emission trends depending on the socio-economic and political situation in each country. A large reduction appears between 1990 and 1995 ( Fig. 4) due to the disintegration of the Soviet Union in 1991. As a result of the economic recession, the reduction in NO x emissions from the power plants and the indus-10 try was twice as large in the east as in the west, despite the introduction of specific abatement measures in the latter.
Except for the effect of this outstanding political situation, the NO x emission trend over the period 1990 to 2005 has been dominated by changes in road transport. However, there are large differences between the east and the west. In Western Europe, 15 road transport has been the dominant NO x emission source over the whole period 1980 to 2005, while power plants were the most important NO x source in Eastern Europe until 1995. For example, the road transport contribution in 1990 varied from less than 20% of total NO x in Ukraine and Kazakhstan to about 60% in France and Switzerland. As a result, 70% of the total European road transport NO x emissions in 1990 came 20 from Western Europe. In 2005, this share dropped to 63% mainly due to reductions in Western European emissions, but also due to increase in emissions in the recovering economies in Eastern Europe.
Due to the dominant character of road transport emissions, Fig. 5 presents the trends in road transport NO x emission, separately for the periods 1980-1990, 1990-2000 25 and 2000-2005 together with a reference map of emissions in 2005 (upper left map). The legend accompanying the difference maps (−20% to +20%) has been chosen to highlight the main differences, but the percentage differences might in certain cases exceed ±50% in any of the three periods considered. Introduction  1 1980-1990 Road transport emissions in Europe increased by 13% in the period 1980-1990 despite a 10% reduction in Eastern Europe (cold colours in the upper right map of Fig. 5). The reduction in the east is linked to decreased fuel consumption due to income deterioration, as a consequence of the inefficiency in resource allocation (investments) 5 (Gros and Steinherr, 1991). Exceptions to this general picture, where emissions increased, are Armenia, the Republic of Moldova, Hungary, Slovenia, Croatia and Estonia. While the explanation for the increased emissions in Hungary can readily be linked to the increase in gasoline consumption, the situation is not clear for the Former Yugoslav and USSR Republics. Fuel consumption data for individual Former Yugoslav 10 and USSR Republics have not been available to us for the period 1980-1990, thus firm conclusions regarding the reasons behind the apparent increase in emissions cannot be drawn. Based on the rather stable fuel consumption trend in both the Former USSR and Yugoslavia between 1980 and 1990, it is not unlikely that fuel consumption increased in some of these Republics and decreased in other parts of this region. 15 In most of Western Europe and Turkey, road transport emissions increased between 1980 and 1990 (warm colours in the upper right map of Fig. 5). The overall increase was 27% in Western Europe. Fuel consumption went down or stabilized also in this region due to the high oil prices following the oil crisis in the 1970s. At the same time, early non-catalyst controls introduced with the different steps of UNECE Regulation No. 15 (1970-1983) were associated with an increase in NO x emissions from vehicles (Berg, 2003). Due to relatively slow fleet turnover, as further discussed in the next section, the introduction of the ECE-R15 regulation maybe responsible for the overall increase between 1980 and 1990. In some Western European countries though, road transport emissions decreased between 1980 and 1990. These are Sweden, 25 Belgium, Luxembourg, Austria, Switzerland, Cyprus and Malta. The fuel consumption went down also in these counties. A possible explanation for the emission decrease could be the early introduction of diesel passenger cars. In the case of Turkey, the Interactive Discussion increased emissions are due to a substantial (80%) increase in gasoline consumption.

1990-2000
In the period 1990-2000, road transport emissions decreased by 23% in Europe, and reductions were evident both in the east and the west (Fig. 5, lower left map). In Eastern Europe, the decrease in emissions is associated with a decrease in fuel consump-5 tion in former Soviet republics, Romania and Bulgaria. A country's transport volume is closely linked to its GDP, and the overall decrease in road transport is an effect of the restructuring of the economies after the disruption of the Soviet Union in 1991. Russia is an important trade partner, so the depression also affected countries outside the Union. Further, the infrastructure in this region which was already rather poor further decayed during this period (EEA, 2007). Decreased emissions in other Eastern European countries are linked to decreased emission factors, rather than decreased fuel consumption. The share of the high-polluting car fleet built in Eastern Europe decreased in these areas between 1990 and 2000 as the increase in the stock of vehicles is due to imports of cleaner cars from Western Europe. This development took place 15 also in the Czech Republic, but here the increased share of lower NO x emitting cars only damped the increase in emissions. Albania and The Former Yugoslav Republic of Macedonia increase their emissions due to increased fuel consumption, without an accompanying decrease in emission factors. The 7% decrease in road transport emissions reported by Croatia cannot be explained without assuming a decrease in 20 emission factors, as both GAINS and IEA report increased fuel consumption in the transportation sector in Croatia between 1990 and 2000. In Western Europe, the introduction of improved vehicle technologies and stringent inspection systems related to the Euro standards has been the primary force in reducing NO x road traffic emissions in the period 1990-2000, despite economic growth 25 and increases in fuel consumption. All countries but Portugal, Spain, Greece, Turkey, Cyprus, Malta, Austria, Ireland and Luxembourg reduced their emissions (Fig. 5  can be divided in three groups based on the possible causes for the emission growth. The high age of the vehicle fleet combined with increasing number of vehicles may explain the lack of reductions in the first group, containing Portugal, Spain and Greece. In the second group with Turkey, Cyprus and Malta the main reason for emission increase is that the Euro standards were not applied at the same time as in the rest 5 of Europe. Increase in emissions reported from Austria, Ireland and Luxembourg are caused by fuel tourism as defined in the previous sections. Austria and Ireland provide road transport emissions both according to fuel sold and fuel used. Their estimates for NO x emissions calculated on the basis of fuel used show a decreasing trend between 1990 and 2000, opposing the data reported as requested by the UNECE Guidelines 10 according to fuel sold. The reason for the increased emission in Austria is a large increase in emissions from heavy duty vehicles (Anderl et al., 2007). In Ireland, the reason is that fuel is less expensive in Ireland than in the United Kingdom during this period. UK fuel prices apply to Northern Ireland, thus drivers tank in Ireland (DEHLG, 2006).

2000-2005
In the period 2000-2005, road transport emissions in Europe continue to decrease by 11%, albeit less than in the preceding decade (23%). Fuel consumption in the traffic sector increased in all European countries except in Germany. In Germany high tax on fuel combined with improvements in vehicle technology, result in a considerable 20 decline in diesel consumption as further discussed in Sect. 5. The situation from the 1980s (Fig. 5 upper right map) with decreasing emissions in the east and increased emissions in west is reversed in this period (Fig. 5 lower right map). Increase in emissions from Eastern Europe follows the increase in fuel consumption ( Fig. 5 lower right map) (EEA, 2007). The EECCA countries have their own car industry, so new western technologies will not necessarily become standard. Another reason why emissions in EECCA countries increase may be related to use of lead as an additive to the fuel. Lead additives destroy the catalysts, and are not completely abandoned due both to lack of regulations and to a claimed black market for leaded gasoline. In addition to this, the price of fuel is low and even subsidised in some countries (EEA, 2007). In Belarus, emission decreased between 2000 and There is no essential production of cars except for heavy duty vehicles here. Produced lorries comply with Euro 2 and later standards and passenger cars are imported. Import of cars which not comply with certain Euro standards are not directly prohibited, but the older cars are imposed

Effectiveness of policy regulations in the transport sector
As indicated in Figs. 3 and 4 The analysis here investigates to what extent emission factors calculated on the basis of officially reported road transport emissions and activity data (implied emission factors) comply with the Euro standards. This information is only available from 1990 to 2005 for ten Western European countries (Austria, Denmark, France, Germany, Netherlands, Norway, Portugal, Spain, Switzerland and United Kingdom) and can be retrieved from the EMEP database (http://webdab.emep.int). These ten countries represent more than 50% of total European emission from road transport in 1990, and they are considered to represent the situation in Western Europe. For Eastern Europe, relevant data are available for 2005 for seven countries (Estonia, Lithuania, Macedonia, Poland, Romania, Slovakia and Slovenia). Therefore, a separate analysis has also 15 been undertaken for this particular year.

Trends in emissions and fuel consumption by country and vehicle class
The fuel types considered here are gasoline and diesel. The reported consumption of hydrogen is negligible and also the reported consumption of compressed natural gas (CNG) and liquefied natural gas (LNG) is very low to affect the trends. In addition, 20 the fuel consumption in mopeds and motorcycles is small compared to other vehicles categories and therefore these have not been considered in the following analysis. It is a limitation to our analysis that the reported emissions do not distinguish between gasoline and diesel use in vehicles. The implied emission factors are thus calculated based on total emissions and fuel consumption for each vehicle class. We argue that 25 our conclusions are robust, albeit we cannot calculate IEFs separately for PCs and LDVs. The reason is that, although the reduction efficiency for NO x in gasoline PC is quite high even for the first generation catalysts, there were relatively few catalysts 10722 Introduction Interactive Discussion installed in 1990. This implies that we can assume emission factors to be higher for gasoline than diesel PC over most of the period 1990 to 2005. Thus our combined IEFs can be considerer lower than the IEFs for gasoline vehicles alone. In Western Europe, the overall trend in fuel consumption shows that petrol consumption decreased (20%), while diesel consumption increased (90%) between 1990 and 5 2005. The shift to diesel is the impact of the European Automobile Manufacturers Association's commitment on the reduction of CO 2 emissions from passenger cars (Commission Recommendation 1999/125/EC) (ACEA, 2007). This agreement promoted the use of diesel passenger cars because they have up to 30% higher fuel efficiency than gasoline cars of similar size. The net fuel consumption in road transport increased about 23% from 1990 to 2005. Passenger cars consumed by far the largest share of fuel (60%), followed by heavy duty vehicles (28%) and LDVs (12%). The promotion of diesel cars via the ACEA Commitment greatly benefited the curtailment of greenhouse gases. At the same time, it should not be forgotten that diesel passenger cars emit as much as three times higher NO x emissions per kilometre than gasoline cars of the 15 same emission standard. Just to put it into perspective, assuming that the increase in fuel consumption would have originated from increase in petrol rather than diesel consumption (thus diesel consumption remaining at the 1990 levels), this would have led to some 1/3 lower NO x emissions in 2005.
With respect to road transport emissions of NO x , these decreased by 44% between 20 1990 and 2005 in Western Europe, despite the increase in fuel consumption. The emission reductions were largest for PC (63%) followed by HDV (21%) and LDV (2%). Figure 6 compares the trends in total fuel consumption from 1990 to 2005 with individual trends for gasoline and diesel consumption and with the total NO x emissions for passenger cars in selected Western European countries. NO x PC emissions decrease For HDV the situation is more complex. Fuel consumption increased in all countries between 1990 and 2005, except in Germany, where HDV consumption decreased by 30% between 2000 and 2005. This substantial decrease in diesel sold is not likely due to technological developments alone, but also due to the high tax on diesel in Germany. The high fuel prices in Germany prevent transit traffic refuelling, and pro-5 mote fuel tourism to other neighbouring countries. Emissions from HDV between 1990 and 2005 decreased in all countries, except in Spain and Portugal and Austria, where emissions increased, by more than 200% in the case of Austria. Austria is a counter case to Germany, in that some 30% of the diesel sold is consumed outside the country. The onset of emission reduction from HDV comes almost ten years later than the cor-10 responding turning point of passenger cars emission. The main reason for the delay in HDV emission reduction is the inefficiency of Euro II standards in addressing NO x . With regard to LDVs, their fuel consumption increased in all countries. Their emission levels have remained relatively stable compared to the emission trends in PC and HDV, with slight increases or decreases in equally many countries. The above results show 15 that the implementation of Euro standards has contributed to a decoupling of emissions and fuel consumption of all vehicle classes in Western Europe since 1990.

Trends in implied emission factors
We have derived implied emission factors (IEF) between 1990 and 2005 based directly on reviewed officially reported emissions and total (gasoline plus diesel) fuel consump-20 tion. In this way we can compare the average emission level of the whole fleet in each country, with the emission levels expected when developing the Euro standards. The results for Western Europe are presented in Fig. 7, which shows that the implied emission factors decrease for all vehicle classes from 1990 to 2005. The average IEF reductions for all countries examined in this period are 67%, 42% and 35% for PC, 25 LDV, and HDV respectively. The periods with largest IEF reductions vary with vehicle class and country. For PC, the largest IEF reductions (35%) occurred between 1995 and 2000 while max reductions for HDVs (20%) appeared five years later ( to 2005). On average, the IEF reductions from LDVs remained relatively constant, at 17%, in all five-year periods. We know today that the introduction of electronic controls in Euro II (1997), and less so in Euro III (2001) heavy duty engines led to excessive NO x emissions over operation modes that were not included in the type-approval test (Hausberger and Rexeis, 2004). As a result, countries with a fast turnover of their HDV 5 fleet were delayed in meeting the stringent emission standards expected. This is the reason that the mean HDV-fleet emission factors in several countries (UK, Netherlands, Austria, Denmark, Switzerland) in 2000 still appears higher than the Euro I emission standard introduced eight years before (1992). The situation improves in 2005 with only Austria and Denmark appearing to have HDV emission levels clearly beyond the emission standards 8 years ago (Euro II).
The conclusions related to the effectiveness of the Euro emission standards for passenger cars are less straightforward, since the IEF is a composite value of gasoline and diesel vehicle emission levels, while separate emission standards have been in place, depending on the fuel used. In general, Fig. 7 shows that the average fleet emission 15 level in 2000 in several European countries corresponded to a level between the gasoline and diesel Euro 1 levels, eight years ago. Spain, with a rather old vehicle fleet, fails to meet the Euro 1 emission standard level even eight years after the introduction of the standard. Even in 2005, the average emissions of the Spanish PC fleet are only marginally below the diesel 1992 levels. On the other hand, average PC fleet emis-20 sions in Germany and Switzerland seem quickly (e.g. within five years) to attain the emission standards, despite the large fleet of diesel passenger cars. A similar decadal delay in effective implementation of the Euro standards is also seen for LDVs.
Due to lack of reported data, changes over time in IEF cannot be determined for Eastern European countries. The analysis for Eastern Europe is restricted to year 25 2005. Table 5 compares implied emission factors for PC, HDV and LDV in Eastern and Western European countries. IEFs in Eastern European countries are a factor of 2.3 and 1.4 higher than in Western European countries for PC and LDV respectively. This is because the Euro standards did not fully apply to Eastern European countries Interactive Discussion before their accession to the European Union in 2004. The result further implies that the implementation of technological measures to abate road transport emissions has been less effective in Eastern Europe, due to a slower turnover of vehicles towards more modern, less polluting technologies. The conclusion from this analysis is that Euro standards have clearly facilitated a 5 substantial reduction in the road transport emissions in Western Europe. There has however taken some eight to ten years for policy regulations to come fully in effect in several European countries. This is the effect of the slow turnover of the vehicle fleet in many countries but also due to the deviations between the emissions in real-world conditions, compared to the type-approval driving cycle. Life cycle assessments of car 10 fabrication might shed light to whether or not the policy should increase the incitement to a faster vehicle turn-over.

Conclusions
The significant increase of liquid fuel consumption in Europe between 1950 and 1980 led to an unparalleled historic increase of NO x emissions from road transport by a 15 factor 14 (Fig. 3). Road transport emissions have been the main source of NO x in Europe already since the 1970s and are currently responsible for about 40% of total anthropogenic emissions. Technological and policy developments to abate European emissions have clearly facilitated a substantial reduction in the NO x levels. Between 1990 and 2005 emissions decreased by more than 30% (Fig. 4, Table 2), but have now 20 started to increase in many Eastern European recovering economies. Based on the development in road transport emissions, we determined five NO x emission trend regimes in Europe. While the emission trends in the first two of these (1880-1950 and 1950-1980) are mainly determined by the development in fuel consumption, the last three regimes are driven more by policy developments. In the third 25 regime (1980-1990) road transport emissions decreased in Eastern Europe and increased in Western Europe (Fig. 5 upper right map). These regional trends resulted in an overall increase in NO x emissions. The emission decrease in Eastern Europe was linked to deterioration in incomes, followed by a decrease in fuel consumption. In Western Europe, fuel consumption increased despite historical high oil prices in the beginning of the period. In addition, the ECE-R15 regulations introduced in the 1970s to improve combustion in motor vehicles, increased NO x emission factors and 5 NO x emissions peaked in 1990. In the fourth regime (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000), large reductions in transport emissions took place all over Europe (Fig. 5 lower left map). In Eastern Europe, decrease in emissions is linked both to decline in fuel consumption in Former Soviet Republics, and to a reduced share of high NO x emitting vehicles in other Eastern European countries. The energy consumption increased in Western Europe, 10 but policy regulations fostered technological development and implementation of measures, which resulted in large decreases in road transport emissions. In the fifth regime (2000)(2001)(2002)(2003)(2004)(2005), the emissions pattern from the 1980s was reversed (Fig. 5  The emission limits for vehicle exhaust introduced by the Euro standards (Table 4) led to a substantial decrease in emission factors for all vehicle types in the EU region (Fig. 7). Despite a significant increase in the fuel consumption over the whole period 1990-2005, emissions monotonically decreased in the case of passenger cars (Fig. 6).
There is clear evidence that the policy approach taken to reduce NO x emissions has 5 been effective in bringing NO x levels down. On the other hand, our study shows broadly in line with Zachariadis et al. (2001) that it takes roughly ten years or more after the introduction of an emission standard to reach an equal level of average fleet emissions. This delay shows one of the inherent limitations with regard to the effectiveness of road transport policy. Although each new emission standard may introduce significant NO x reductions over the one it replaces, it takes several years before a substantial portion of the fleet complies with the new emission standard. This leads to a rather gradual reduction in emissions from road transport compared to stationary sources, where new regulations have a more instant effect after their implementation. In order to fully account for the delay in compliance, it is important to note that introduction of new 15 technologies was in some cases accompanied by HDVs and passenger cars emitting much higher in real-world operation than the emission standard level (Hausberger and Rexeis, 2004;Ntziachristos and Samaras, 2000), due to loopholes in the type-approval procedure. This analysis shows that diesel consumption in vehicles increased substantially be- 20 tween 1990 and 2005. For the purpose of abating CO 2 emissions, this development is very welcomed, but from the air quality perspective, the "dieselization" hampered a more rapid NO x abatement. If the increase in fuel consumption since 1990 in passenger cars had been met by an increase in gasoline rather than diesel, this would have resulted in around 30% lower NO x emissions today (2005). The above considerations 25 led us to conclude that the policy aimed at reducing NO x from the transport sector has not been as effective as the ambition level. Some issues for future considerations are identified. Due to the increase in diesel consumption, primary NO 2 emissions may be increasing in Europe, despite the over- Interactive Discussion all reduction in NO x . In diesel exhausts, excess oxygen may lead to much higher NO 2 /NO x ratios (50%) in vehicles equipped with oxidation aftertreatment (diesel oxidation catalyst or catalyzed filter) for PM control (AQEG, 2006) than for gasoline three-way catalytic converter cars (less than 5%). Current evidence shows that ambient concentrations of NO 2 do not decrease at the same rate as NO x in various European hotspots 5 (Lambrecht, 2007;Carslaw et al. 2007), mainly due to the increasing NO 2 ratio in late diesel technology vehicles. Hourly NO 2 concentration limit values become mandatory in Europe starting from 2010 (EC Daughter Directive 99/30/EC). The proportion of primary NO 2 in vehicle exhausts may need to be addressed in future NO x inventories. Another effect of the increase in diesel consumption in road transport is that less nonmethane volatile organic pollutants (NMVOC) emissions are emitted from this sector. The average NO x /VOC emission ratio for PC and LDV has increased by a factor 2 between 1990 and 2005 according data officially reported to the UNECE. The impact on tropospheric ozone production of the above EU wide changes in emission ratios from road transport should be further assessed by air quality modelling. 15 It has been demonstrated here that implied emissions factors for NO x in 2005 are sometimes even higher than the emission standard requested 15 years before. Although the slow vehicle replacement rate is responsible for a large part of this deviation, an equally significant part originates from the discrepancy of real-world operation emissions and emission standards. The next stage of emission standards expected 20 in Europe by 2009-2010 (Euro 5) will further reduce NO x emissions from both gasoline and diesel vehicles. This will be achieved with both in-cylinder measures and aftertreatment devices, such as selective catalytic reduction (SCR) systems and lean-NO x catalysts. It needs to be made sure that these devices will be effective over the complete or, at least, a large portion of the engine operation range, to avoid exces-25 sive off-cycle emissions. Emission control regulations in the future should therefore more effectively address off-cycle emissions, e.g. by introducing a type-approval test covering a wider range of engine operation modes.
This paper does not analyse the implications for NO x emissions from the transport