Health impact assessment of PM10 and EC in the city of Rotterdam (the Netherlands) in the period

Transcription

1 1 / 40 Health impact assessment of PM10 and EC in the city of Rotterdam (the Netherlands) in the period M.P. Keuken and P. Zandveld (TNO), S. van den Elshout (DCMR), G. Hoek (IRAS University of Utrecht), N. Janssen (RIVM Bilthoven) September 2010 March 2011 TNO-060-UT This report is a publication in the framework of the Netherlands Research Program on Particulate Matter II (BOP II) performed by the Energy research Centre of the Netherlands (ECN), the Environment and Safety Division of the National Institute for Public Health and the Environment (RIVM) and TNO. The research of BOP II is supervised by a steering committee which consists of Menno Keuken (TNO), Ronald Hoogerbrugge (RIVM), Ernie Weijers (ECN), Eric van der Swaluw (RIVM), Klaas Krijgsheld (Ministry of Infrastructure and the Environment) and Jan Matthijsen (PBL Netherlands Environmental Assessment Agency).

2 2 / 40

3 3 / 40 Summary In the framework of the Netherlands Policy Support Program on PM ( BOPII ), the trend of PM 10 and elemental carbon (EC) concentrations and related health impact has been assessed in the city of Rotterdam in the period EC is regarded an indicator for exhaust emissions of combustion aerosol. In , international, national and local measures have been implemented to reduce in particular exhaust emissions of combustion aerosol by road traffic. However, the annual growth of road traffic kilometers in the same period was 1.5% on inner-urban roads and 3% at motorways in the Netherlands. Despite the volume growth of road traffic, it is concluded that after 1995, road traffic emissions of PM 10 and EC on urban roads have decreased, while on non-urban roads these emissions remained constant. Since 2003 monitoring data of EC show no decreasing trend which suggests that exhaust emissions by the growing traffic volume are no longer compensated by cleaner vehicles. Our study shows, that in Rotterdam in the period the air quality of PM 10 and EC improved significantly both at urban background and near heavy traffic locations. This results in a gain in life years on average of 13±6 months (PM 10 ) or 12±8 months (EC) for the population in Rotterdam. The health impact for PM 10 and EC is similar though the population weighted PM 10 concentration dropped on average 18 µg.m -3, while for EC this was 2 µg.m -3. From research in the Netherlands it is concluded that 70% of the decrease in PM 10 concentrations the last decades is related to secondary inorganic aerosol and only for 10% to primary PM emissions, including combustion aerosol. The similarity in health impact for PM 10 and EC suggests that the health impact of PM 10 is mainly related to the contribution of combustion aerosol in PM 10 and less to the contribution of secondary inorganic aerosol. This demonstrates that EC is a more sensitive indicator (compared to PM 10 ) to monitor the health effects of traffic measures. Also, it is concluded that measures directed to reduce combustion aerosol (e.g. exhaust emissions of road traffic and (inland) shipping) are more effective to reduce health effects of air quality than reducing PM 10 in general. It is noted, that EC is likely not causing the health effects but acts as a proxy for the mass of combustion aerosol. Further experimental research is recommended to improve modeling of EC in urban areas (e.g. establish emission factors of EC for free-flowing and congested road traffic) and to validate effects of traffic measures on air quality of EC (e.g. low emission zones and 80 km/h speed limitation). On-line measurements of EC at TNO s dynamometer test facilities demonstrate that the MAAP instrument is an appropriate instrument to establish emission factors for EC.

4 4 / 40

5 5 / 40 Samenvatting In het Beleidsondersteunend Programma voor Fijnstof ( BOPII ) is onderzoek gedaan naar de trend van PM 10 en elementair koolstof (EC), en gerelateerde gezondheidseffecten in de periode in Rotterdam. EC is een indicator voor de uitstoot van verbrandingsaerosol door (diesel)verkeer. In eerdergenoemde periode was de jaarlijkse groei in voertuigkilometers 1.5% voor binnenstedelijk verkeer en 3% op snelwegen in Nederland. Ondanks deze groei van het wegverkeer zijn door schonere voertuigen de verkeersemissies van PM 10 and EC vooral na 1995 op binnenstedelijke wegen verminderd, terwijl op buitenstedelijke wegen deze emissies gelijk bleven. Sinds 2003 dalen de EC concentraties niet meer, wat suggereert dat uitlaatemissies door het toenemende wegverkeer niet meer worden gecompenseerd door schonere voertuigen. Onze studie laat zien dat in Rotterdam in de periode , de luchtkwaliteit van PM 10 en EC significant is verbeterd op de stedelijke achtergrond en de buurt van druk wegverkeer. Dit levert een winst in levensjaren van gemiddeld 13±6 maanden door PM 10 of 12±8 maanden door EC per persoon in Rotterdam. De gezondheidseffecten van PM 10 en EC zijn in dezelfde ordegrootte, terwijl de bevolkingsgewogen concentratie van PM 10 met gemiddeld 18 µg.m -3 daalde en van EC met slechts 2 µg.m -3. Onderzoek in BOPI Nederland liet zien dat 70% van de afname in PM 10 concentraties de afgelopen decennia wordt toegeschreven aan secondair anorganisch aerosol en voor 10% aan primaire PM emissies met o.a. verbrandingsaerosol. De overeenkomst in gezondheidseffecten voor PM 10 en EC wijst erop dat gezondheidseffecten van PM 10 in Rotterdam vooral zijn toe te schrijven aan het aandeel verbrandingsaerosol. Dit laat zien dat EC een gevoeliger indicator is (vergeleken met PM 10 ) om gezondheidseffecten van verkeersmaatregelen te beoordelen. Tevens wordt geconcludeerd dat beleid gericht op het terugdringen van verbrandingsaerosol, zoals het verminderen van uitlaatemissies door (diesel)verkeer en scheepvaart, effectiever is voor het verminderen van gezondheidsrisico s door luchtvervuiling dan het generiek verlagen van PM 10 concentraties. Het wordt benadrukt dat EC waarschijnlijk niet de oorzaak is van de gezondheidseffecten maar een indicator voor de verspreiding van de massa van verbrandingsaerosol. Experimenteel vervolgonderzoek wordt aanbevolen om verspreidingsmodellen voor EC in stedelijk gebied te verbeteren. Verder is het gewenst om het effect van verkeersmaatregelen zoals emissiezones en doorstromingsmaatregelen (o.a. 80 km/u op snelwegen en groene golven op binnenstedelijke wegen ) met metingen te valideren. Verkennende on-line metingen met de MAAP van EC in uitlaatemissies geeft aan dat de MAAP een veelbelovend instrument is voor het vaststellen van EC emissiefactoren van wegverkeer.

6 6 / 40

7 7 / 40 Contents Summary... 3 Samenvatting Introduction Objective Study approach Experimental approach Modeling Sampling period and locations Monitoring instruments Results Limit values for PM vehicular exhaust emissions in Actual emission factors for PM and EC in the Netherlands in Actual traffic emissions for PM and EC in the Netherlands in Air quality in Rotterdam in Air quality and health effects in Rotterdam in Conclusions and recommendations References Appendices A Exhaust emission measurements of EC and PM

8 8 / 40

9 9 / 40 1 Introduction Exposure to elevated levels of particulate matter (PM) has been associated with health effects in epidemiological surveys. Most of these studies focused on the mass of PM without distinction of specific sources [Anderson, 2009]. However, in urban areas the population is specifically exposed to traffic-related PM emissions by road traffic [Künzli et al., 2000]. This concerns a complex mixture of particles from tire wear, break/friction processes, re-suspension of road dust and exhaust. Typically, exhaust emissions include a large number of submicron particles ( combustion aerosol ) which have been associated with health effects [Obersdörster et al., 2005; Aalto et al., 2005; Sioutas et al., 2005; Laden et al., 2000; Lanki et al., 2006; Fischer et al., 2009]. Also, these particles contribute to global warming [Novakov and Hansen, 2004]. Combustion aerosol consists of a large number of particles smaller than 100 nm ( ultrafine particles ), while most of the mass is found in larger particles [Kittelson et al., 2006; Maricq, 2007]. Ultrafine particles are primarily formed from condensation of hot semi-volatile hydrocarbons within a few seconds in the exhaust wake [Sakurai et al., 2003; Kumar et al., 2009]. The larger particles originate from the combustion chamber and are mainly composed of elemental carbon (EC) and condensed organic compounds (OC). These larger particles seem better conserved during dispersion, while ultrafine particles are more reactive [Zhang et al., 2004a and 2004b]. Over the years, a variety of international and local measures have been implemented to reduce exhaust emissions by road traffic. Examples are more stringent emission standards and environmental zoning in cities to prevent high emitters near residential areas. Consequently, there is a need to assess the impact of these measures on air quality and health. However, even near heavy traffic locations, PM 10 and PM 2.5 are not adequate indicators, as they are dominated by regional background concentrations [Zhu et al., 2002]. The concentration of elemental carbon (EC) is considered a more appropriate indicator for dispersion and exposure of the population to traffic-related PM [Schauer et al., 2003]. EC may be measured by thermal analysis of PM collected on a filter [Chow et al., 2001 and 2009; Cavalli and Putaud, 2007]. The thermal method with optical transmittance correction of artifact EC formation from OC is expected to become the reference method for EC [Birch and Cary, 1996]. Emission factors for road traffic of EC have been established based on thermal analysis [Ntziachristos and Samaras, 2009] and concentration response functions for health effects of EC have been established [Janssen et al., 2011]. Therefore, in this report EC is applied as an indicator to study the effects of traffic emissions on air quality and health over the period in the city of Rotterdam (the Netherlands). Black Smoke measurements and conversion to EC were applied to establish EC time series. Black Smoke has been measured as part of the national and regional air quality monitoring networks at the urban and regional background in and near Rotterdam, the Netherlands. The Black Smoke method is based on optical measurement of the reflectance of PM sampled on a filter, which is converted to mass units by the Black Smoke Index [ISO 9835, 1993]. Since the introduction of this index in the 1950s, the relation between reflectance and mass has changed drastically [Quincey, 2007]. As a result, the Black Smoke mass units overestimate the mass of combustion aerosol as measured by thermal analysis. In addition, the

10 10 / 40 Black Smoke method suffers from internal reflectance in the filter material and interferences by white (e.g. secondary inorganic aerosols, sea salt and soil) and brown (e.g. organic compounds) particles collected on the filter. It was noted by Quincey (2007), that perhaps the greatest limitation when re-evaluating historical data of Black Smoke is the lack of knowledge on the exact methods and quality assurance applied at the time, for example filter types and flow calibrations. However, in Rotterdam Black Smoke measurements at the regional location and an urban traffic location were performed by respectively the National Institute for Health and Environment (RIVM) and the Regional Environmental Agency (DCMR) both following the same ISO procedure as part of the national and regional air quality monitoring network under stringent QA/QC accreditation. This assured good quality over the whole period from From 2003 onwards, the manual Black Smoke measurements in the National Air Quality Monitoring Network were exchanged with automatic measurements by the SX200 (ETL, Hereford, UK). This automated method provides similar results as the manual method following the ISO 9835 protocol [Hijink, 2002]. The aim of this study is to assess the health impact of PM 10 and EC in the city of Rotterdam in the period The population in Rotterdam in this period was about in 1985 and in 2008 (COS, 2009). Rotterdam has the largest harbour of Europe and consequently, road traffic is intensive with a relatively large contribution of heavy duty vehicles. The spatial distribution of PM 10 and EC is computed by dispersion modelling [Denby et al., 2011; Air4EU, 2005]. In this study, the URBIS model is used which is a combination of a street canyon model and a line-source model for motorways [Beelen et al., 2010]. The modelled PM 10 and EC concentrations are compared with results in 2008 from the regional air quality network in Rotterdam (DCMR EPA Rijnmond). The trend in PM 10 and EC provides information on the trend in air quality in The exposure of the population to PM 10 and EC over this period is computed as well as the health impact. In section 2 and 3 the objective and the approach of the study are detailed. In section 4 the experimental setup is described and in section 5, the results are presented and discussed. The conclusions and recommendations are elaborated in section 6. The study was financed by the Ministry of Infrastructure and Environment in the framework of the Netherlands Policy Support Program on Particulate Matter ( BOPII ).

11 11 / 40 2 Objective The objective in this study is to assess the health impact of PM 10 and EC in the period in the city of Rotterdam.

12 12 / 40

13 13 / 40 3 Study approach The URBIS model is applied to estimate the spatial distribution of PM 10 and EC in the city of Rotterdam for the years 1985, 1995 and The spatial resolution of the URBIS model is a 10*10 m 2 grid from 10 m from the road side up to 500 m near motorways and up to the housing façade for inner-urban roads. The spatial distribution of annual average PM 10 and EC for 1985, 1995 and 2008 was combined with GIS based population distribution data. This provides information on the population exposure to long-term air pollution at house address. The population data for 2008 were also applied for 1985 and 1995 to eliminate the effects of variation in the population on the health impact assessment. In the next step, concentration-response-function for long-term health effects of PM 10 and EC are combined with the population exposure maps. Finally, the health impact assessment is used to evaluate the trend in health effects in the period in relation to exposure to PM 10 and EC.

14 14 / 40

15 15 / 40 4 Experimental approach 4.1 Modeling The required model input concerns: meteorological conditions; As model input, the ten year average meteorological conditions for the period 1995 to 2004 was applied for the years 1985, 1995 and Identical meteorological conditions were used to eliminate the effect of meteorological variation on the health impact assessment; traffic data; For traffic in Rotterdam in the years 1985 and 1995, the number of vehicles was calculated from the number of vehicles in the base year 2008 with a decreasing annual trend of previous years of 3% for motorways and 1.5 % for inner-urban roads. For the years 1995 and 1985, the traffic composition of private cars, middle duty and heavy duty vehicles in the year 2008 was applied; emission factors; Emission factors of PM related to road traffic (e.g. both wear and exhaust emissions) for the period 1990 to 2008 are available for the car fleet in the Netherlands [PBL, 2009]. These emission factors were used to extrapolate emission factors for For EC, the emission factors were derived from an EU database with EC emission factors as a fraction of PM exhaust emissions [Ntziachristos and Samaras, 2009]. For PM 10, the exhaust emission factors were augmented with non-exhaust emissions by friction processes (e.g. brakes and tires). This increased the exhaust emission factors with 10 mg.km -1 for private cars, 50 mg.km -1 for middle duty vehicles and 70 mg.km -1 for heavy duty vehicles [PBL, 2009]; regional and urban background; For 2008, the background was based on the 1*1 km 2 national concentration map of PM 10 in the Netherlands [PBL, 2009]. For the previous years 1995 and 1985, the background was computed from an increasing trend of 0.7 µg PM 10 per m 3 per year starting in 2008 [Matthijssen and Koelemeijer, 2010]. The regional and urban backgrounds of EC were derived by conversion of Black Smoke measurements in and near Rotterdam in the period The conversion of Black Smoke to EC concentrations by a factor 10 was established by data from parallel measurements of Black Smoke and EC near a motorway A10 in Amsterdam [Keuken et al., 2010] and various sites in Rotterdam [Beker, 2006]. For the year 2008, the spatial distribution of PM 10 and EC in the city of Rotterdam as modelled by the URBIS model was validated by measurements at an urban background and traffic locations.

16 16 / Sampling period and locations In the period , Black Smoke measurements have been performed as part of national and local monitoring networks at a regional background location Westmaas and two urban, traffic locations Floreslaan and Vasteland in Rotterdam. Westmaas is a regional background location 15 km south of Rotterdam. The locations are presented in Figure Figure 1: Relevant sampling locations in and near Rotterdam: Westmaas (regional), 433 Floreslaan (urban traffic) Vasteland (urban traffic) and Schiedam (urban). 4.3 Monitoring instruments The MAAP (Thermoscience Model 5012) is an automatic monitoring instrument for EC based on measurements of both the transmission and reflectance of light (660 nm) by particulate matter collected on filter tape [Petzold and Schönlinner, 2004]. The MAAP converts its optical measurements into EC mass units by a default calibration based on the German VDI thermal protocol. This may result in an overestimation of 15-30% as compared to other thermal protocols such as the EUSAARII protocol. The MAAP has been employed by DCMR to measure continuously EC concentrations at urban and traffic sites in Rotterdam. The MAAP has been used as it is able to provide hourly measurements and because thermal analysis of EC is too laborious to apply in routine monitoring networks.

17 LDV (mg.km -1 ) and HDV (mg.kwh -1 ) 17 / 40 5 Results 5.1 Limit values for PM vehicular exhaust emissions in The mass of exhaust emissions has been regulated by EU emission limit values since The trend in mass based emission limits for vehicles in the EU from 1992 onwards is illustrated in Figure 2. EU emission limit (PM) LDV HDV (Euro: 1 / I) 1996 (Euro: 2 / II) 2000 (Euro: 3 / III) 2005 (Euro: 4 / IV) 2011/2008 (Euro: 5 / V) 2014/2013 (Euro: 6 / VI) Figure 2: EU limit values for mass based emission factors for exhaust PM in diesel light duty vehicles (LDV) in mg.km -1 and heavy duty vehicles (HDV) in mg.kwh -1 in the period 1992 till 2014 for Euro 1-6 (LDV) and Euro I-VI (HDV). In Figure 2, LDV refers to cars with a weight lower than 3.5 ton (passenger cars and vans), while HDV relates to heavy duty vehicles with a weight over 3.5 ton. For gasoline-fueled LDV only from 2011 onwards, emission limits for PM have been regulated in Europe. Figure 1 illustrates that in the period 1992 till 2005, PM emission limits for vehicles have decreased a factor 6 for LDV: from 140 to 25 mg.km -1 and a factor 18 for heavy duty vehicles: from 360 to 20 mg.kwh -1. These limit values relate to the total mass in exhaust emissions without distinction in elemental carbon and organic compounds 1. The more stringent emission requirements over the last two decades reflect improved engine technology and fuel composition (e.g. the sulfur content in diesel fuel reduced from 350 ppm in 2000 to sulfur free diesel with less than 10 ppm in 2009). 1 EURO 5/6 emission standards - effective from 2011/2014 for passenger gasoline and diesel vehicles also include a limit for the number of particles larger than 23 nm of 6 * particles per km. This number-based limit was introduced to prevent that vehicles may meet the mass limit but allow emissions of high numbers of particles. Still, this new EU emission standards leaves particles smaller than 30 nm unregulated.

18 Emission factor PM (mg.km -1 ) 18 / Actual emission factors for PM and EC in the Netherlands in Actual emission factors of vehicles are investigated by dynamometer testing. In the Netherlands, each year a number of in-use and new vehicles are selected for exhaust emission tests. This concerns specific driving cycles to simulate typical driving conditions at different road types. The results of these measurements are converted into mass based emission factors in accordance to the COPERT guidelines [Ntziachristos and Samaras, 2000]. Emission factors may also take into account non-exhaust emissions of particulates as a result of friction processes by tires and brakes. Re-suspension of road dust by road traffic is not included in emission factors. Non-exhaust emissions by friction processes and re-suspension become relatively more important in the future. These emissions remain constant while exhaust emissions are decreasing. The net result is a decreasing trend in the contribution of exhaust emissions in emission factors: 90% in 1990, 80% in 2000 and 45% in Emission factors of road traffic for PM in the Netherlands for the period are presented in Figure LDV urban LDV motorway HDV urban HDV motorway Figure 3: Emission factors of PM for free-flowing traffic on urban roads and motorways for LDV and HDV in the period in the Netherlands [Source: PBL, 2009]. The data for 1985 in Figure 3 were extrapolated from the data, as no PM emission factors were available previous Figure 3 illustrates that for LDV (passenger cars and vans) the actual PM emission factors for urban areas have indeed decreased in the period by a factor 6, as required under the emission limit values set by the EU. For heavy duty vehicles, the actual reduction in emission factors in this period for motorways was a factor 6 which is less than the intended reduction by a factor 18 (see Figure 2). The fraction of elemental carbon in exhaust emissions have been derived from dynamometer testing and are presented in Table 1 [Ntziachristos and Samaras, 2009].

19 19 / 40 Table 1: Fraction EC of PM exhaust emissions of gasoline/diesel light duty vehicles (LDV) and diesel high duty vehicles (HDV) in the period before 1992 ( pre-eu standard and conventional ) and from 1992 till 2014 for Euro 1-6 (LDV) and Euro I-VI (HDV). Category Euro standard EC/PM (%) Gasoline-LDV Pre-EU standard Euro Euro Euro Euro Diesel-LDV Conventional Euro Euro Euro Euro Euro Diesel-HDV Conventional Euro I Euro II Euro III Euro IV Euro VI Uncertainty (%) 1 : PM emissions for gasoline LDV were not regulated under Euro 1-4. Only from 2011, gasoline LDV have to comply with a limit value for PM in exhaust emissions under Euro 5. 2 : The EC fraction for pre-euro standards were estimated by Ntziachristos and Samaras (2009) for gasoline- LDV, diesel-ldv and HDV as respectively 2%, 55% and 50% which is considerable lower than the Euro standards. However, a study in the USA [EPA, 2008] concluded that no model year or age dependency on the EC/PM ratio in period was measured. Therefore, in our study the pre-euro standards are based on an average of the estimation by Ntziachristos and Samaras (2009) and the Euro exhaust emissions. Table 1 shows that for Euro 1-4 and Euro I-IV the fraction of EC in PM exhaust emissions was relatively constant and on average 20% (gasoline LDV), 80% (diesel LDV) and 70% (HDV). Our measurements of the EC/PM ratio in exhaust emissions of a Euro-III HDV simulating free-flowing driving at a highway, resulted in 64% (see: Annex 1). This figure is in good agreement with the value in Table 1 of 70% for a NEDC-cycle (e.g. cold start, acceleration, urban driving, motorway) resulting in a higher EC fraction. For the period before 1992, the EC fraction is significantly lower. For the future, it is expected that after introduction of Euro 5 and Euro VI, the total mass of PM in exhaust emissions will be reduced drastically (see: Figure 2) as well as the EC fraction for all vehicles. Road tunnel and road side measurements resulted in a fraction EC of PM 2.5 in the range of 20-70% depending on the amount of heavy duty vehicles [Gratmonev et al., 2003; Ning et al. 2008; Naser at al. 2009]. These experimental data agree with the dynamometer testing. Therefore, it is concluded that the ratio EC/PM presented in Table 1 is a valid basis to estimate EC emission factors from PM emission factors for road traffic. The results are presented in Table 2.

20 20 / 40 Table 2: Emission factors of PM and EC (mg.km -1 ) in exhaust emissions of road traffic on urban roads and motorways in the Netherlands in the period PM 1 (mg.km -1 ) EC/PM 2 (%) EC (mg.km -1 ) urban motorway urban motorway 1985 LDV (20% diesel 3 ) Diesel-MDV Diesel-HDV LDV (30% diesel) Diesel-MDV Diesel-HDV LDV (35% diesel) Diesel-MDV Diesel-HDV LDV (50% diesel) Diesel-MDV Diesel-HDV : Source PBL; 2 : Table 1; 3 % diesel from total number of transport kilometers by LDV in the Netherlands. 5.3 Actual traffic emissions for PM and EC in the Netherlands in The total PM emissions by road traffic are computed from the product of the emission factors and the total number of vehicle kilometers in the Netherlands. The number of kilometers driven by all vehicles in the Netherlands in the period is presented in Figure 4.

21 billion vehicle kilometers (km) 21 / 40 Traffic performance in the Netherlands gasoline-pc diesel-pc MDV HDV Figure 4: The number of vehicle kilometers by gasoline and diesel passenger cars, middle duty vehicles (MDV) and heavy duty truck (HDV) in the Netherlands in the period [source: CBS: Figure 4 shows that the number of vehicle kilometers for passenger cars have increased by about 30% in the period 1985 to 2008 in the Netherlands. For diesel passenger cars and middle duty vehicles the number of vehicle kilometers has tripled in this period, while for heavy duty trucks the number of vehicle kilometers almost doubled. The impact on exhaust emissions of PM and EC by the growth of road traffic in the Netherlands is calculated by using the emission factors (see: Table 2) and vehicle kilometers (see: Figure 4). The results for the period are presented separately for urban and non-urban areas in Figure 5. The fractions of vehicle kilometers on urban/non-urban roads are based on data for 2005, which are for private cars and trucks, respectively 25/75% and 10/90%.

22 Mkg 22 / 40 Annual exhaust emissions in urban and rural areas (the Netherlands) PM10-urb PM10-rur EC-urb EC-rur Figure 5: Exhaust emissions expressed as PM 10 and EC by road traffic on urban ( urb ) and non-urban ( rur ) roads in the Netherlands in million kg per year in the period Figure 5 illustrates that in the period : 1.) the total exhaust emissions of PM 10 and EC by road traffic have been reduced by respectively 30% and 0% and 2.) traffic emission on urban roads have been reduced almost by 50%. From Figure 4, it is concluded that in the period , PM 10 and EC emissions at nonurban roads (e.g. motorways) have increased as a result of volume growth (see: Figure 4) followed by decreasing emissions till 2000 due to cleaner vehicles. After 2000, on urban roads traffic emissions still show a decreasing trend but on nonurban roads emissions remain constant. Hence, cleaner vehicles do not longer balance the volume growth of road traffic at non-urban roads. More than 80% of population in the Netherlands lives in urban areas with relatively high exposure to road traffic emissions. The effects of road traffic emissions have been investigated on urban air quality and associated health effects in the city of Rotterdam. 5.4 Air quality in Rotterdam in Air quality of PM 10 in Rotterdam in The annual average concentrations for PM 10 in Rotterdam were modelled for the years 1985, 1995 and 2008 by the URBIS model (see Section 4.1). The modelling results of the spatial distribution for PM 10 are presented in Figure 6A-C. In these figures, the motorways and the river Oude Maas with harbour areas in the west are presented in black.

23 23 / 40 Figure 6A: Annual average concentrations of PM 10 (µg.m -3 ) in Rotterdam (1985). Figure 6B: Similar as Figure 5A for the year 1995.

24 24 / 40 Figure 6C: Similar as Figure 5A for the year Figures 6A-C illustrate that the air quality for PM 10 in the city of Rotterdam has improved significantly. The urban background for PM 10 decreased from 43 µg.m -3 in 1985 to 25 µg.m -3 in The value of 43 µg.m -3 for 1985 is in good agreement with the annual average of 48 µg.m -3 PM 10 in the Netherlands estimated by the National Environmental Protection Agency (RIVM) [Knol and Staatsen, 2005]. The decrease in PM 10 concentrations between 1985 and 2008 is attributed for 70% by large-scale emission reductions by industry, energy production and road traffic of precursors (sulfur dioxide, ammonia and nitrogen oxides) of secondary particles [Hoogerbrugge et al., 2010]. The remaining 30% reduction of PM 10 is related to primary PM emissions (10%), secondary organic aerosols (10%) and less water adsorbed to particulate matter (10%). Figures 6A-C show that in 1985 PM 10 concentrations were elevated more than 20% near inner-urban roads and motorways. From 1995 onwards, the contribution of primary road traffic emissions to local air quality decreased to less than 10%. Hence, despite the annual growth of traffic, the air quality of PM 10 in the period improved significantly near heavy traffic locations. This improvement is mainly the result of decreasing background concentrations and only limited by lower exhaust emissions of primary PM Air quality of EC in Rotterdam in To perform a similar analysis for EC as for PM 10, the trend in background concentrations of EC was estimated from Black Smoke measurements. Time series of Black Smoke measurements are available for the regional background near Rotterdam ( Westmaas ) and the urban, traffic locations Floreslaan and

25 Annual Black Smoke Index ( mg.m -3 ) 25 / 40 Vasteland (see: Figure 1). The time series of Black Smoke measurements for the period are presented in Figure 7. Rotterdam: regional traffic 1 traffic Figure 7: Annual average BS index (µg.m -3 ) at traffic locations 1 ( Floreslaan ) and 2 ( Vasteland ) in Rotterdam and a regional background location near Rotterdam in the period Figure 7 shows that the delta between the two traffic locations and the regional background of black smoke decreased in the period As the traffic locations are more exposed to urban traffic emissions than the regional background, this indicates that the reduction in urban traffic (exhaust) emissions were higher than other (large-scale) sources of combustion aerosol. Since 2003 no further decreasing trend in the delta can be detected and this indicates that reduced emissions by cleaner vehicles do not longer compensate the growth in traffic volume. To evaluate the trend in combustion aerosol by EC, the Black Smoke data have been converted into EC concentrations. The relation between Black Smoke and EC has been investigated by parallel measurements of Black Smoke and thermal analysis of EC by the GGD-Amsterdam using the NIOSH protocol near a motorway A10 in Amsterdam and at various locations in Rotterdam in the period The results are presented in Figure 8A-B.

26 BS index (mg.m -3 ) BS index (mg.m -3 ) 26 / 40 Motorway A10 - Amsterdam: (10:1) EC (mg.m -3 ) Figure 8A: Scatter plot of BS index (µg.m -3 ) and EC (µg.m -3 ) in 24-h samples collected near the motorway A10 in Amsterdam in 2006 and 2007 (n=111) [Source GGD Amsterdam]. Rotterdam: (10:1) regional urban traffic EC (mg.m -3 ) Figure 8B: Scatter plot of BS index (µg.m -3 ) and EC (µg.m -3 ) in 24-h samples collected at rural, urban and traffic sites in and near Rotterdam in the period November 2006 till August 2007 (n=67). Figures 8A-B illustrate that the Black Smoke index is linear correlated with EC near a motorway with more than vehicles per day near Amsterdam and at various regional, urban and traffic locations in and near Rotterdam. The statistical parameters for the motorway in Amsterdam and the locations in Rotterdam are respectively for the slopes 9 and 11, the intercepts near zero and the regression coefficients (R 2 ) 0.8 and 0.9. Based on these results, it is concluded that EC concentrations may be derived from Black Smoke measurements by division with a factor 10.

27 27 / 40 This relation has been used to estimate from the Black Smoke measurements (see: Figure 7) the regional/urban background in Rotterdam of EC: 2.6/3.2 µg.m -3 (1985), 1.1/2.6 µg.m -3 (1995) and 0.7/1.1 µg.m -3 (2008). The meteorological conditions and traffic input to calculate the contribution of exhaust emissions tot EC concentrations in 1985, 1995 and 2008 in Rotterdam were similar as applied for PM 10 in section Emission factors for EC by road traffic were based on the data presented in Table 2. The results are presented in Figure 9A-C. Figure 9A: Annual average concentrations of EC (µg.m -3 ) in Rotterdam (1985). Figure 9B: Similar as Figure 7A for the year 1995.

28 28 / 40 Figure 9C: Similar as Figure 7A for the year Figures 9A-C demonstrate that similar to PM 10 the air quality for EC improved significantly in the period This is attributed to lower regional background concentrations as a result of reduced emissions of soot particles by combustion processes in general (e.g. industry, energy production) and at urban scale of diesel-fueled road traffic in particular. Also, near inner-urban roads and motorways, the air quality of EC improved due to lower emissions of especially, diesel-fuelled road traffic. Consequently, the contribution of road traffic near heavy traffic decreased from 1-2 µg.m -3 in 1985 to µg.m -3 in Hence, despite the annual growth of road traffic, the air quality of EC has improved in the period Validation of modelled air quality in Rotterdam in 2008 Comparison between the concentrations calculated by the URBIS model and measurements for the year 2008 for PM 10 and EC are presented in Table 3. The model calculations were performed with meteorological data for the year 2008 and not the ten-year average as applied for the trend analysis in section The measurements were performed by monitoring stations of the regional environmental protection agency DCMR in Rotterdam.

29 29 / 40 Table 3: Measurements and modelling results of the annual average PM 10 and EC at locations in Rotterdam (2008). PM 10 (µg m -3 ) EC (µg m -3 ) URBIS Monitoring URBIS monitoring Urban background - Schiedam A C Traffic location - Floreslaan A D - Vasteland 28.9 n.a D Motorway station - Ridderkerk B 2.2 n.a. A : gravimetric; B : Tapered element oscillating monitor TEOM corrected with 1.3; C : Multi angle absorption photometry (MAAP); D : Conversion from Black Smoke index; n.a.: not available The uncertainty in monitoring annual concentrations is in the order of 15%, while for modelling, the uncertainty is in the range of 25 to 40 % for an urban background and road side location, respectively [PBL, 2009]. Considering these uncertainties, the results in Table 4 show good agreement between modelled and monitoring annual averages for PM 10. For EC, the modelling and monitoring results at the urban background location differ relatively more than PM 10. This may be explained by a.) underestimation of the modelled urban background by too large conversion factor 11 of Black Smoke to EC concentrations and b.) overestimation of the measured background by the MAAP instrument of EC concentrations due to calibration of MAAP by VDI thermal EC analysis by 30% [Keuken et al., 2010]. Our study underlines that more experimental data on EC is required to further improve modelling of dispersion EC in urban areas. 5.5 Air quality and health effects in Rotterdam in Population density in Rotterdam in The population in Rotterdam has been estimated from information on zip codes per X,Y-coordinate in Zip codes contain a mixture of households, offices and shops. The average number of persons per household in Rotterdam is estimated at 2.0 [COS, 2009]. On the basis of this calculation, we have identified inhabitants. This estimate agrees with the official population figure of in 2008 and in To investigate the health impact assessment in , we have maintained a constant figure for the population of Rotterdam at The number of inhabitants per X,Y-coordinate have been combined with the air quality for PM 10 and EC in the period (see: section 5.4). The exposure of the population in Rotterdam has been classified to various levels of PM 10 and EC and presented in Table 4.

30 30 / 40 Table 4: The number of inhabitants in Rotterdam exposed to various levels of annual average PM 10 and EC in the period Number of inhabitants (#) PM 10 (µg.m -3 ) EC (µg.m -3 ) Health effects of PM 10 and EC in Rotterdam in Epidemiological studies have provided evidence of associations between concentrations of PM and several adverse health outcomes including: cardiorespiratory mortality, hospital admissions for cardiovascular and respiratory disease, asthma attacks, acute bronchitis, respiratory symptoms and restriction in activity [Brunekreef, 2002; WHO, 2006]. In general, a health impact assessment (HIA) of outdoor air pollution by PM 10 and EC is based on four components [Ostro, 2004]: 1. an assessment of the ambient air concentrations of PM 10 and EC by monitoring or model-based estimation; 2. a determination of the size of the population exposed to specific concentrations of PM 10 and EC; 3. a determination of the health effect of prime interest, including the baseline rate of the health effect being estimated (e.g. the underlying mortality rate in the population in deaths per thousand people); 4. a derivation and application of concentration-response functions from the epidemiological literature that relate ambient concentrations of PM 10 and EC to selected health effects. Population exposure distributions (steps 1 and 2) were taken from Table 4. Health impact calculations were performed for the midpoints of the exposure categories and a rounded value just below the lowest category and above the highest category, see Table 6 for the exact values. Though air pollution has been associated with both mortality and morbidity effects, quantitatively the effects of mortality have been shown to be most important in previous health impact assessments (e.g. Künzli, 2000). We therefore focus on quantification of mortality effects. Mortality effects of long-term exposure are substantially larger than mortality effects related to short-term daily exposures [Künzli, 2000; WHO, 2006]. We therefore derived an exposure response function based upon long-term exposure studies.

31 31 / 40 Exposure response functions were selected from a recent review of the evidence for PM 2.5 and EC [Janssen et al., 2011]. For PM 2.5, we used RR (95% confidence interval ) expressed per 1 µg.m -3. For EC we used RR 1.06 (95% confidence interval ) expressed per 1 µg.m -3. Note that concentration levels of PM 2.5 exceed those of EC substantially. We assumed that we could apply the PM 2.5 exposure response function to the Rotterdam case, even though exposure was characterized as PM 10. This may be problematic for the calculation of the health impact for a particular year, but much less so for the calculation of differences between years as almost all decrease in PM 10 is due to a decrease of the fine fraction of PM 10 [Hoogerbrugge et al., 2010]. We have expressed mortality impacts in life years gained or lost estimated with life table calculations [Miller and Hurley, 2003]. For the calculation we used a population of 500,000 people aged 18 to 64, distributed in age categories comparable to the 2008 Dutch population [CBS, 2008]. We have estimated the effects on this population for a lifetime, as follows: For 1985, 1995 and 2008 we first calculated the life years lost related to the exposure distribution in that year. We then subtracted the life years lost in 1995 and 2008 from the life years lost in 1985 to calculate the gain in life years related to a decrease in concentration. To assess the uncertainty, these calculations were performed for the best estimate of the exposure response functions and the 95% lower and upper confidence interval. In order to asses the effect of a decrease in concentration, we used the same population size and same baseline mortality rates for the calculations in the three years. The results of the calculations are presented in Table 5 (best estimate of exposure response function) and Table 6 (reflecting uncertainty in exposure response function). The decrease in PM 10 concentration from 1985 to 2008 results in a gain in life of on average 13 months per person. For EC, a gain of 12 months per person is calculated. Table 6 illustrates that the uncertainty in the exposure response function has a fairly large impact on the estimated gain in life years. The range of the estimated gain in life between 1995 and 2008 based upon PM10 was 3 to 15 months, around a mean estimate of 9 months (Table 6). This is in good agreement with a study in 2008 which estimated a gain in life years of 6 months between 1995 and 2006 computed as 20% in increased life expectancy of 2.3 years is attributed to improved air quality, including particulate matter [PBL, 2008]. The health impact is similar for PM 10 and EC, which is not in agreement with an assessment made of a local traffic reduction policy which showed much large health benefits when the calculation was based upon EC [Janssen et al., 2011]. The difference between this and the Janssen study is the relative decrease in concentration. For example, the resulted decrease of PM 10 in an environmental zone which reduces exhaust emissions is mainly attributed to a decrease in EC. Hence, the health benefits based on calculations for EC show much higher benefits as a result of the ten times higher RR for EC as compared to PM 10. Contrary, in the current study, the population weighted PM 10 concentration dropped from 43.2 µg.m - 3 in 1985 to 24.8 µg.m -3 in 2008, while EC only dropped from 3.1 to 1 µg/m 3 over the same time period. From research in the Netherlands it is concluded that 70% of the decrease in PM 10 concentrations the last decades is related to secondary inorganic aerosol and only for 10% to primary PM emissions, including combustion aerosol. The similarity in health impact for PM 10 and EC suggests that the health impact of

32 32 / 40 PM 10 is mainly related to the contribution of combustion aerosol in PM 10 and less to the contribution of secondary inorganic aerosol. Table 5: Loss of life years in Rotterdam computed from the exposure to annual average PM 10 and EC in the period Loss of life years PM 10 (µg.m -3 ) (25) a (32.5) (37.5) (42.5) (50) Total (years) Total per person (months) b Change compared to 1985 (months) b 4 13 EC (µg.m -3 ) (1.0) (2.0) (3.0) (4.0) (5.0) Total (years) Total per person (months) b Change compared to 1985 (months) b a In brackets the number used for the life year calculations. Population size ( inhabitants), age distribution and baseline mortality rates kept constant (2008 mortality rates used, absolute number from table 5). b Calculated by dividing the total life years lost by the total population (570,000 inhabitants)

33 33 / 40 Table 6: Uncertainty of loss of life years in Rotterdam computed from the exposure to annual average PM 10 or EC in the period , using lower and upper limit of the exposure response relationship for two years (1995 and 2008) PM 10 (µg.m -3 ) (25) a (32.5) (37.5) (42.5) (50) 0 0 Total (years) Total per person (months) b Change compared to 1995 (months) b 3 15 EC (µg.m -3 ) (1.0) (2.0) (3.0) (4.0) (5.0) 0 0 Total (years) Total per person (months) b Change compared to 1995 (months) b 4 20 a In brackets the number used for the life year calculations. Population size ( inhabitants), age distribution and baseline mortality rates kept constant (2008 mortality rates used, absolute number from table 5). b Calculated by dividing the total life years lost by the total population (570,000 inhabitants)

34 34 / 40

35 35 / 40 6 Conclusions and recommendations In this study, PM 10 and EC have been applied to study the effect of traffic emissions on air quality and health over the period in the city of Rotterdam. In the Netherland, the number of vehicle kilometers for passenger cars have increased by about 30% in the period 1985 to For diesel passenger cars and middle duty vehicles the number of vehicle kilometers has tripled in this period, while for heavy duty trucks the number of vehicle kilometers almost doubled. In the period , total exhaust emissions of PM 10 and EC by road traffic have been reduced in the Netherlands by respectively 30% and 0% and on urban roads emissions have been reduced almost by 50%. From monitoring data it is concluded that after 2003 cleaner vehicles do not longer compensate the volume growth of road traffic. To evaluate the trend in combustion aerosol by EC, the Black Smoke data have been converted into EC concentrations. A conversion factor 10 was derived from research in Amsterdam and Rotterdam. With this factor, urban background concentrations for EC were computed for 1985, 1995 and Emission factors for EC were computed as fraction of PM exhaust emissions for personal cars, middle duty trucks and heavy duty trucks respectively 10%, 65% and 60% (before 1992) and 20%, 80% and 70% (after 1992). Our study shows that in Rotterdam in the period the air quality of PM 10 and EC improved significantly both at urban background and near heavy traffic locations. This results in a gain in life on average of 13±6 months (PM 10 ) or 12±8 months (EC) per person in Rotterdam. The health impacts for PM 10 and EC are similar though the population weighted PM 10 concentration dropped on average 18 µg.m -3, while for EC this was only 2 µg.m -3. The ten times larger drop in PM 10 concentrations as compared to EC results in similar health impact. This demonstrates that EC is a more sensitive indicator (compared to PM 10 ) to monitor the health effects of traffic measures. Also, it is concluded that measures directed to reduce combustion aerosol (e.g. exhaust emissions of road traffic and (inland) shipping) are more effective to reduce health effects of air quality than reducing PM 10 in general. It is noted, that EC is likely not causing the health effects but acts as a proxy for the mass of combustion aerosol. Further experimental research is recommended to improve modelling of EC in urban areas (e.g. establish emission factors of EC for free-flowing and congested road traffic) and to validate effects of traffic measures on air quality of EC (e.g. low emission zones and 80 km/h speed limitation). The MAAP instrument is regarded a promising instrument to establish on-line emission factors for EC.

36 36 / 40

37 37 / 40 7 References 1. Aalto P., Hämeri K., Paatero P., Kulmala M., Bellander T. and Berglind N. (2005) Aerosol particle number concentration measurements in five European cities using TSI-3022 condensation particle counter over a threeyear period during health effects of air pollution on susceptible subpopulations. Journal of Air and Waste Management 55: Anderson H.R. (2009) Air pollution and mortality: a history. Atmospheric Environment 43: Air4EU (2005). Final recommendations on AQ assessment Beelen R., Voogt M., Duyzer J., Zandveld P. and Hoek G. (2010). Comparison of the performance of land use regression modeling and dispersion modeling in estimating small-scale variations in long-term air pollution concentrations in a Dutch urban area. Atmospheric Environment 44: Beker, D. (2006). Relation between black smoke and MAAP 5012 measurements in Dutch. DCMR, Rotterdam, the Netherlands 6. Birch M.E. and Cary R.A. (1996). Elemental carbon-based method for monitoring occupational exposures to particulate diesel exhaust. Aerosol Science and Technology 25: Brunekreef B. and Holgate S.T. (2002). Air pollution and health. Lancet 360: CBS (2008). Central Office for Statistics Hoogerbrugge R., Denier van der Gon H.A.C., van Zanten M.C. and Matthijsen J. (2010). Trends in particulate matter. BOP-report / Cavalli F. and Putaud J-P (2007). Towards a standardized thermal-optical protocol for measuring atmospheric organic and elemental carbon; the EUSAAR protocol Chow J.C., Watson J.G., Crow D., Lowenthal D.H. and Merrifield T. (2001). Comparison of IMPROVE and NIOSH carbon measurements. Aerosol Science and Technology 34: Chow J.C., Watson J.G., Doraiswamy P., Antony Chen L-W, Sodeman D.A., Lowenthal D.H., Park K., Arnott W.P. and Motallebi N. (2009). Aerosol light absorption, black carbon, and elemental carbon at the Fresno Supersite, California. Atmospheric Environment 93: Centrum voor Onderzoek en Statistiek (2009). Population Rotterdam (in Dutch). project