- Email: [email protected]
Trends in hydrocarbon fleet emissions at four UK highway sites
The Science of the Total Environment 235 Ž1999. 91]99
Trends in hydrocarbon fleet emissions at four UK highway sites D.M. RevittU , G.M. Muncaster, R.S. Hamilton Urban Pollution Research Centre, Middlesex Uni¨ ersity, Bounds Green Road, London, N11 2NQ, UK
Abstract The on-road hydrocarbon emissions of vehicles at four sites in the UK have been monitored using remote sensing equipment. The single lane highway sites in London ŽHaringey and Southwark., Middlesborough and Leicester were monitored between May 1994 and October 1995. Analysis of the results shows that there is both a large majority of low emitting vehicles which contribute little to fleet hydrocarbon emissions and a small minority of high emitting vehicles which contribute significant proportions to fleet hydrocarbon emissions at all sites. This also results in a skewing of the data set so that a pattern of high mean values and lower median values is consistently observed. Analysis of model year data suggests a low association between vehicle age and mean hydrocarbon emissions for vehicles produced prior to 1983 but the relationship improves after 1983 with regression analyses giving r 2 values as high as 0.96. Relatively new high polluting vehicles are the greatest contributors to fleet emissions with, on average, 52% of hydrocarbon fleet emissions being produced by these vehicles from model years 1985]1991. Therefore, fleet emissions could be significantly reduced if new highly polluting vehicles were subject to regular emissions testing followed by appropriate remedial action or were removed from the highway by the withdrawal of their vehicle registration. Older vehicles play a minor role in fleet emissions with, on average, only 13% of hydrocarbon fleet emissions being produced by vehicles registered prior to 1983. Q 1999 Elsevier Science B.V. All rights reserved.
1. Introduction Road transport systems can influence the lifestyles of local populations through a variety of deleterious effects including deterioration in air quality, noise disturbance, disruption to communities, destruction of existing urban and rural U
Corresponding author.
environments, depletion of resources, and contamination of soils, watercourses and groundwaters. In particular, the exhaust emissions associated with increasing road traffic levels throughout the UK are leading to increased health and environmental problems. Carbon monoxide ŽCO., carbon dioxide ŽCO 2 ., nitrogen oxides ŽNO x . and hydrocarbons ŽHC. are the major gaseous air pollutants produced by petrol-engined vehicles and the measurement of the emissions of two of
0048-9697r99r$ - see front matter Q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 8 - 9 6 9 7 Ž 9 9 . 0 0 1 9 5 - 3
92
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
these Žcarbon monoxide and hydrocarbons. is incorporated into the annual test carried out on all vehicles over 3 years of age in the UK. However, a more realistic assessment of the emission characteristics of vehicles in relation to type, make and age can be achieved by using on-road emissions monitoring particularly if any pre-warning to the vehicle owners can be avoided. The Fuel Efficiency Automobile Test ŽFEAT. was developed in 1987 to initially measure CO and CO 2 exhaust emissions from on-road vehicles, as an aid to the estimation of fleet fuel consumption ŽStedman et al., 1991c. but has since been extended for the monitoring of HC, NO x and smoke. A qualitative assessment of hydrocarbons in vehicular exhaust emissions is obtained from the hydrocarbon to carbon dioxide ŽHCrCO2 . ratio with a high ratio being indicative of a highly polluting vehicle and conversely a low ratio representing a clean vehicle with respect to hydrocarbon emissions. The FEAT remote sensor has been validated against on-board measurements with agreements to within "15% for hydrocarbons and "5% for carbon monoxide ŽAshbaugh et al., 1992.. Comparison of measurements made using different remote sensors for the same emissions provided r 2 values of between 0.92 and 0.99 for CO and values of between 0.77 and 0.87 for HC ŽStephens and Cadle, 1991.. Comparison of FEAT emission measurements with idle emission measurements has been less successful due to the dissimilarity in vehicle operating modes and speeds. However, quite a high degree of agreement was observed between FEAT results and a conventional system operated on a chassis dynamometer ŽHickman and McCrae, 1995.. Zhang et al. Ž1995. have reviewed the data collected from the use of the FEAT system to monitor vehicle exhaust emissions at 22 locations around the world. A strong relationship was observed between mean fleet emissions for CO and mean fleet emissions for HC due to common influencing factors such as differing levels of maintenance, average fleet age and imposition of compulsory emission standards. However, an individual vehicle that is a high emitter of CO is not necessarily a high emitter of HC. Quintile analy-
sis has been conducted on emissions data from four cities ŽGothenburg, Los Angeles, Melbourne and Leicester . and the higher CO and HC emissions for the highest emitting quintile of the newest vehicles compared with the lowest emitting quintile of the oldest vehicles indicates that in-use emission differences between well maintained and badly maintained vehicles is larger than the age-dependent deterioration of emissions.
2. Methodology and site characteristics 2.1. The FEAT system The FEAT system utilises standard spectroscopic principles and combustion equations to measure the CO and HC emissions from passing vehicles in under 1 s ŽBishop et al., 1989.. It consists of four major components: an infrared source, a detector, a computer and a video camera. A horizontal beam of infrared radiation is directed across a single lane of traffic approximately 30 cm above the road surface. As a vehicle breaks this infrared beam, measurements are taken of the infrared absorption both in front of the vehicle and as the vehicle exits the beam. The detector converts the incident infrared radiation to a voltage signal and the COrCO 2 and HCrCO2 ratios are measured to compensate for dilution effects which occur as the exhaust plume moves away from the vehicle. These ratios can be computed to the percentage exhaust composition of CO and HC. The system requires daily calibration with a certified gas mixture of CO, CO 2 and propane. The use of propane means that the hydrocarbon results are expressed in terms of ‘propane equivalents’. The video camera, which records 60 framesrs, is focused on the rear of vehicles as they pass the sensor and the image of each vehicle is recorded together with the date, time, and the percentage of CO, HC and CO 2 in the emissions. The video information is recorded on tape and if required, the vehicle can be identified according to model, make and age by comparison with the Drivers Vehicle Licensing Authority ŽDVLA. records.
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
93
Table 1 Characteristics of monitored sites Site
Name
Sampling date
Traffic flow rate Žvehiclesrh.
Vehicle speeds
No of valid hydrocarbon readings
A
Bounds Green Road, Haringey Dixons Bank, Middlesborough Abbey Road, Southwark Uppingham Road, Leicester
May 1994
511]780
8]32
June 1994
465]650
40]80
March and August 1995
127]376
4]50
October 1995
315]701
8]30
11 099 Ž85%. 3979 Ž53%. 7414 Ž74%. 12 223 Ž81%.
B C D
2.2. Monitoring site characteristics The four monitored sites were all single lane carriageways. The relevant important characteristics of each site including the ranges of traffic flow rates, the ranges of vehicle speeds and the number of valid HC readings are described in Table 1. The low number of valid HC readings at Site B is due to the unfortunate loss of hydrocarbon channel data for one complete day.
3. Discussion 3.1. Fleet emissions of hydrocarbons Analysis of the measured hydrocarbon emissions according to their percentage contribution
to the fleet and to the total fleet emissions shows that the majority of vehicles are low emitters and contribute little to total fleet emissions. Thus, at site A ŽHaringey., 69% of vehicles ŽHC emission range - 1000 ppm. contributed to only 24% of fleet emissions ŽFig. 1.. Sites B ŽMiddlesborough., C ŽSouthwark. and D ŽLeicester . produced similar results ŽTable 2.. Those vehicles in the emission category representing hydrocarbon concentrations of less than 1000 ppm in Middlesborough, Southwark and Leicester, accounted for 67, 75 and 65% of the fleets, respectively, but contributed only 5, 18 and 9%, respectively, of fleet emissions. Further analysis of data from site A shows that those vehicles emitting - 2000 ppm accounted for 89% of the fleet, but contributed only 50% to total fleet emissions. Further analyses of the data from sites B]D produced a similar trend to that
Table 2 Summary of hydrocarbon emission statistics Site
A B C D
Mean Žppm.
1000 1510 940 1552
Median Žppm.
590 420 350 450
Emission category - 1000 ppm
Emission category - 2000 ppm
Emission category ) 10 000 ppm
Vehicles Ž%.
Fleet emissions Ž%.
Vehicles Ž%.
Fleet emissions Ž%.
Vehicles Ž%.
Fleet emissions Ž%.
69 67 75 65
24 5 18 9
89 82 88 79
50 19 35 23
1 3 1 4
14 40 19 33
94
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
Fig. 1. Hydrocarbon emissions for site A ŽBounds Green Road, Haringey. expressed as percentage contribution to the fleet and the fleet emissions.
of site A with a large proportion of the fleet contributing comparatively little to fleet emissions ŽTable 2.. The converse situation is also true in that a small minority of highly polluting vehicles Žhydrocarbon emission concentrations ) 10 000 ppm. contribute a significant proportion to fleet emissions, with 3% of the fleet at site B producing 40% of fleet emissions. Site D produced a very similar result with 4% of the fleet producing 33% of total fleet emissions. However, sites A and C although agreeing with the general trend, only had 1% of the fleet in the emission concentration range ) 10 000 ppm which contributed 14 and 19% of the fleet emissions, respectively. This situation, although agreeing with the general pattern found for CO emissions ŽMuncaster et al., 1996., differs in that a greater proportion of the total fleet emissions are contributed by the highly polluting vehicles.
The mean emissions for sites A]D are 1000, 1510, 940 and 1552 ppm, respectively. The corresponding median values at each site are 590, 420, 350 and 450 ppm ŽTable 2.. The differences between the mean and median values reflect the very skewed nature of the emission distributions ŽFig. 1.. Calculation of the skewness values for the emission distributions shows site C to have the most skewed distribution with a skewness value of 2.52 although similar values were found at all sites Žsite A, 2.17; site B, 2.28; site D, 2.29.. Similar distributions and skewness values have been observed for CO at these sites ŽMuncaster et al., 1996.. The emission distributions are similar to those seen in other remote sensing studies carried out in the UK ŽVanke and Bidgood, 1992; Muncaster et al., 1994. and overseas ŽStedman et al., 1991b; Sjodin, 1994.. The mean emission level obtained for Leicester
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
is lower than the 2100 ppm value recorded by Zhang et al. Ž1995. for the same city but similar to the 1650 ppm found by Vanke and Bidgood Ž1992.. The 2100 ppm value would compare Leicester with Mexico City, Kathmandu and Bangkok in the hierarchical cluster analysis performed by Zhang et al. Ž1995.. The fleets in these cities are composed of older vehicles that are less well maintained compared to Leicester and not subject to emission control legislation or new emission standards. The results from this study are more realistic in that they place Leicester in direct comparison with cities where the fleets are newer, better maintained and are subject to emission control legislation. The mean emission levels for the two London sites are similar but are lower than the 1400 ppm found by Zhang et al. Ž1995. and the 1910 ppm obtained by Vanke and Bidgood Ž1992. . The London fleet exhibits similar hydrocarbon emission distributions to those of Edinburgh, Melbourne and Chicago according to Zhang et al. Ž1995. and the results obtained from both the Haringey and Southwark monitoring studies agree with this. In
95
making these comparisons it is important to recognise that remote sensing measurements obtained on roads in the same urban area may be influenced by different factors, such as traffic origin]destinations, which will affect the emission distribution. 3.2. Model year hydrocarbon emissions. The existence of a relationship between mean emissions and the age of the vehicle has been investigated by plotting mean hydrocarbon emissions against model year. The results, including error bars calculated using a 95% confidence interval, show a large degree of scatter with consistently low r 2 values at all sites. The Leicester fleet Žsite D. showed the highest association between model year and mean hydrocarbon emissions with a r 2 value of 0.26. The Haringey Žsite A., Middlesborough Žsite B. and Southwark Žsite C. fleets all gave r 2 values of 0.09, indicating that only 9% of the variation in mean emissions was explained by the model year of the vehicle. A strong influencing factor in this low correlation is
Fig. 2. Relationship between mean annual hydrocarbon emissions, with error bars calculated using a 95% confidence interval, and model years, 1985]1995, for Site D ŽUppingham Road, Leicester . Ž r 2 s 0.96..
96
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
the increased variability with vehicle age which is clearly illustrated by the large error bars on the older model year cars. The small sample sizes of these vehicles is also an important factor and a similar effect has been observed for CO ŽMuncaster, 1996.. Examination of the emissions data from site A shows that out of a total of 7017 valid vehicle measurements, only 912 measurements Ž13% of the fleet. were recorded for those vehicles produced before 1983. The equivalent emissions data statistics for sites B]D were 169 valid measurements Ž4% of the fleet., 350 valid measurements Ž8% of the fleet., 667 valid measurements Ž10% of the fleet., recorded, respectively, for those vehicles manufactured before 1983. For 1965, 1968 and 1969 there was only one vehicle measured for each model year at site A, with only two vehicles measured for years 1967 and 1970. For this reason, only those vehicles registered between 1983 and 1993 for sites A and B and 1985 and 1995 for sites C and D have been included in an additional investigation of the variation of hydrocarbon emissions with model year. The result is a considerable decrease in the size of the error bars and a scatter plot displaying the relationship between mean hydrocarbon levels and model year for site D is displayed in Fig. 2. The Leicester fleet together with that analysed for site C ŽSouthwark. showed the greatest association between the model year of a vehicle and mean pollutant emissions Ž r 2 values of 0.96 and 0.91, respectively.. The regression equations for the relationships found at these sites have similar large negative regression coefficient values of y172.77 and y190.63, respectively. This indicates that at both sites mean pollutant emissions change in a similar pattern such that a small increase in model year is accompanied by a large decrease in pollutant level. This relationship was also found for CO ŽMuncaster et al., 1996.. Sites A and B have similar regression coefficients of y92.52 and y68.8, respectively, so that in comparison to sites C and D, a larger increase in model year is required for the same decrease in pollutant level. This difference may be due to the additional catalyst equipped model years at sites C and D compared to sites A and B. These
additional catalyst-equipped model year vehicles increase the slope of the regression line by accentuating the contrast between the older non-catalyst-equipped vehicles and the more modern technology vehicles such that a small increase in model year is accompanied by a large decrease in pollutant level. The developed regression equations can be used to predict mean hydrocarbon emissions for different model years and these can be compared to the observed values. There is an over-prediction of mean emissions for model year 1993 at all sites which can be attributed to the introduction in January 1993 of EU Directive 91r441rEEC. This Directive introduced new emission limits which were most readily achieved by the fitting of threeway catalysts ŽTWC.. Therefore, a proportion of the hydrocarbon emissions not explained by the regression equations can be attributed to the introduction of catalyst technology to the UK fleet. Mean emissions for model year 1992 are also lower than predicted, and are explained by the early introduction of some catalyst-equipped vehicles ahead of Directive 91r441rEEC. In contrast, model years 1994 and 1995 at site D show an under-prediction of mean emissions which occurs because the steep decline in emissions from older non-catalyst-equipped vehicles to new catalyst-equipped vehicles is not sustainable. Due to all post-1992 model year vehicles having catalysts there is a plateauing of emissions at a very low base level and not a reduction, hence the under-prediction of emissions. However, it is important to note that EU Directive 91r441rEEC was only the latest of many vehicle emission regulations beginning with EU Directive 70r220rEEC Žinstituted to take measures against air pollution from vehicle engine emissions. for which the effect has been to reduce emissions from motor vehicles over time. 3.3. Model year contribution to fleet hydrocarbon emissions Although emissions generally increase with the age of the vehicle, older vehicles do not contribute significantly to fleet emissions due to the small number of old vehicles on the road. Analy-
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
sis of vehicle numbers and fleet emissions at site A shows that pre-1983 registered vehicles contribute only 18% of total fleet emissions. Site B fleet emissions demonstrate a similar scenario with pre-1983 registered vehicles Ž4% of the fleet. contributing only 5% to fleet emissions. Scrutiny of vehicle numbers and fleet emissions at sites C and D displays a similar pattern with only 12% and 16%, respectively, of total fleet emissions, contributed by pre-1983 manufactured vehicles. Because new vehicles are low emitters, it is not true that they do not contribute significantly to total fleet emissions. Analysis of vehicle numbers in relation to fleet emissions at site A shows that those vehicles registered during and after 1983 Ž80% of the fleet. contribute 82% of total fleet emissions. Site B emissions data reveals a similar pattern with vehicles registered during and after 1983 contributing 95% of total fleet emissions. Emissions data for sites C and D are 88% and 84%, respectively, of total fleet emissions being produced by vehicles registered during and after 1983. However, reduction of on road emissions by blanket targeting of this more modern age group would not necessarily produce the desired result. The vast majority of these vehicles are low emitters and it is the gross polluter component that
97
contributes significantly to total fleet emissions and therefore needs specific targeting. A gross polluter is defined as any vehicle that falls within the top 20% of emitters Žquintile 5. at a specific site when the model year is rank ordered. This is illustrated by an analysis of the emissions data from site A ŽFig. 3.. The gross polluter component of model year 1987, a mere 139 vehicles out of a fleet total of 7017 vehicles, contributed 7.7% of total fleet emissions. This pattern was repeated for those vehicles registered in 1988 with the gross polluter component Ž152 vehicles. contributing 8.1% of total fleet emissions. An analysis of the emissions data from site B also illustrates the impact high emitters have on fleet emissions with the gross polluter component of model year 1989 contributing 10.7% of total fleet emissions from just 2.3% of the fleet. A similar pattern was observed at sites C and D with the gross polluter component of model year 1988 at site C Ž1.9% of the fleet. contributing 9.5% of total fleet emissions. For those vehicles registered in 1985 at site D, the gross polluter component Ž1.5% of the fleet. contributed 8.9% of total fleet emissions. Therefore, fleet emissions could be significantly reduced if those vehicles in quintile 5 received regular emissions testingrservicing or were removed from the highway by the withdrawal of
Fig. 3. Hydrocarbon quintile emissions for site A ŽBounds Green Road, Haringey..
98
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
their vehicle registration. An investigation of the site A data shows that if only those vehicles in quintile 5, registered between 1985 and 1991, were targeted then there is the potential to reduce fleet emissions at site A by 41%. The same analysis at sites B]D shows that fleet emissions could be reduced by 66, 51 and 48%, respectively. Rueff Ž1992. conducted a remote sensing survey to see if emission reduction was possible and concluded ‘that the repair of high-emitting vehicles identified via remote sensing constitutes a viable, cost-effective means of reducing certain types of vehicle emissions’. Similar work ŽBishop et al., 1993; Cadle et al., 1993. has shown that after repairs had been carried out emissions had decreased by more that 50% for the repaired vehicles. Older gross polluters are less important because there are so few vehicles in this category and therefore they have a small impact on total fleet emissions. An analysis of emissions from site A shows the minimal importance of older gross polluters ŽFig. 3.. If those vehicles registered in 1976 are examined, it can be shown, that only two vehicles were gross polluters, and they contributed only 0.2% to total fleet emissions. An analysis of emissions from the other sites shows a similar trend. The hydrocarbon quintile emissions distributions are similar in form to those of CO at these sites ŽMuncaster, 1996. and to those seen in other remote sensing surveys ŽStedman et al., 1991a,b; Guenther et al., 1994. with the greater contribution to fleet emissions coming from the gross polluters in quintile 5 with little contribution to fleet emissions from old vehicles.
4. Conclusions The results of this survey indicate that fleet emissions could be reduced markedly if only a small proportion of vehicles are repaired or removed from the road. One possible way of achieving this would be to employ remote sensing techniques to screen the in-use fleet for high emittingrgross polluting vehicles, which would then be subject to further test and repair. Remote
screening has the additional advantage of being performed without warning, thereby identifying vehicles that may otherwise have been tampered with before and after conventional inspection as well as those which attempt to avoid inspections completely. References Ashbaugh LL, Lawson DR, Bishop GA, Guenther PL, Stedman DH, Stephens RD, Groblicki PJ, Parikh JS, Johnson BJ, Huang SC. On-road remote sensing of carbon monoxide and hydrocarbon emissions during several vehicle operating conditions. In: AWMArEPA Conference on PM 10 Standards and Particulate Control. 1992; 9:1]10. Bishop GA, Starkey JR, Ihlenfeldt A, Williams WJ, Stedman DH. IR long path photometry, a remote sensing tool for automobile emissions. Anal Chem 1989;61:671a]677a. Bishop GA, Stedman DH, Peterson JE, Hosick TJ, Guenther PL. A cost-effectiveness study of carbon monoxide emissions reduction utilising remote sensing. J Air Waste Manage Assoc 1993;43:978]988. Cadle SH, Gorse RA, Lawson DR. Real-world vehicle emissions: a summary of the 3rd annual co-ordinating council-air pollution research advisory committee ŽCRC-APRAC.. J Air Waste Manage Assoc 1993;43:1084]1090. Guenther PL, Bishop GA, Peterson JE, Stedman DH. Emissions from 200, 000 vehicles: a remote sensing study. Sci Total Environ 1994;146r147:297]302. Hickman AJ, McCrae IS. Evaluation of a remote vehicle emission measurement system. Transport Research Laboratory Project Report No. 1052, 1995. Muncaster GM, Hamilton RH, Revitt DM, Stedman DH, Vanke J. Individual emissions from on-road vehicles. In: Proceedings of 27 th ISATA Conference on The Motor Vehicle and the Environment, 1994;29:267]274. Muncaster GM. Vehicle emissions and roadside air quality. Unpublished PhD thesis. London, UK: Middlesex University, 1996:282. Muncaster GM, Hamilton RH, Revitt DM. Remote sensing of carbon monoxide vehicle emissions. Sci Total Environ 1996;189r190:149]153. Rueff RM. The cost of reducing emissions from late-model high emitting vehicles detected via remote sensing. J Air Waste Manage Assoc 1992;42:921]925. Sjodin A. On-road emission performance of late-model TWC-cars as measured by remote-sensing. J Air Waste Manage Assoc 1994;44:397]404. Stedman DH, Bishop GA, Peterson JE, Guenther PL. On-road CO remote sensing in the Los Angeles Basin. Contract No. A932-189. Sacramento: CARB, 1991a. Stedman DH, Bishop GA, Peterson JE, Guenther PL, McVey IF, Beaton SP. On-road carbon monoxide and hydrocarbon remote sensing in the Chicago area. Report No. ILENRrRE-AQ 15, 1991b.
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99 Stedman DH, Bishop GA, Pitchford M. Evaluation of a remote sensor for mobile source CO emissions. Report No. EPAr600r4-90r032, 1991c. Stephens P, Cadle SH. Remote sensing of carbon monoxide emissons from road vehicles. J Air Waste Manage Assoc 1991;41:39]43.
99
Vanke J, Bidgood JFS. Remote sensing of vehicle emissions}principles and potential. Inst Mech Eng 1992: C387]C457. Zhang Y, Stedman DH, Bishop GA, Guenther PL, Beaton SP. Worldwide on-road vehicle exhaust emissions study by remote sensing. Environ Sci Technol 1995;29:2286]2294.
Trends in hydrocarbon fleet emissions at four UK highway sites D.M. RevittU , G.M. Muncaster, R.S. Hamilton Urban Pollution Research Centre, Middlesex Uni¨ ersity, Bounds Green Road, London, N11 2NQ, UK
Abstract The on-road hydrocarbon emissions of vehicles at four sites in the UK have been monitored using remote sensing equipment. The single lane highway sites in London ŽHaringey and Southwark., Middlesborough and Leicester were monitored between May 1994 and October 1995. Analysis of the results shows that there is both a large majority of low emitting vehicles which contribute little to fleet hydrocarbon emissions and a small minority of high emitting vehicles which contribute significant proportions to fleet hydrocarbon emissions at all sites. This also results in a skewing of the data set so that a pattern of high mean values and lower median values is consistently observed. Analysis of model year data suggests a low association between vehicle age and mean hydrocarbon emissions for vehicles produced prior to 1983 but the relationship improves after 1983 with regression analyses giving r 2 values as high as 0.96. Relatively new high polluting vehicles are the greatest contributors to fleet emissions with, on average, 52% of hydrocarbon fleet emissions being produced by these vehicles from model years 1985]1991. Therefore, fleet emissions could be significantly reduced if new highly polluting vehicles were subject to regular emissions testing followed by appropriate remedial action or were removed from the highway by the withdrawal of their vehicle registration. Older vehicles play a minor role in fleet emissions with, on average, only 13% of hydrocarbon fleet emissions being produced by vehicles registered prior to 1983. Q 1999 Elsevier Science B.V. All rights reserved.
1. Introduction Road transport systems can influence the lifestyles of local populations through a variety of deleterious effects including deterioration in air quality, noise disturbance, disruption to communities, destruction of existing urban and rural U
Corresponding author.
environments, depletion of resources, and contamination of soils, watercourses and groundwaters. In particular, the exhaust emissions associated with increasing road traffic levels throughout the UK are leading to increased health and environmental problems. Carbon monoxide ŽCO., carbon dioxide ŽCO 2 ., nitrogen oxides ŽNO x . and hydrocarbons ŽHC. are the major gaseous air pollutants produced by petrol-engined vehicles and the measurement of the emissions of two of
0048-9697r99r$ - see front matter Q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 8 - 9 6 9 7 Ž 9 9 . 0 0 1 9 5 - 3
92
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
these Žcarbon monoxide and hydrocarbons. is incorporated into the annual test carried out on all vehicles over 3 years of age in the UK. However, a more realistic assessment of the emission characteristics of vehicles in relation to type, make and age can be achieved by using on-road emissions monitoring particularly if any pre-warning to the vehicle owners can be avoided. The Fuel Efficiency Automobile Test ŽFEAT. was developed in 1987 to initially measure CO and CO 2 exhaust emissions from on-road vehicles, as an aid to the estimation of fleet fuel consumption ŽStedman et al., 1991c. but has since been extended for the monitoring of HC, NO x and smoke. A qualitative assessment of hydrocarbons in vehicular exhaust emissions is obtained from the hydrocarbon to carbon dioxide ŽHCrCO2 . ratio with a high ratio being indicative of a highly polluting vehicle and conversely a low ratio representing a clean vehicle with respect to hydrocarbon emissions. The FEAT remote sensor has been validated against on-board measurements with agreements to within "15% for hydrocarbons and "5% for carbon monoxide ŽAshbaugh et al., 1992.. Comparison of measurements made using different remote sensors for the same emissions provided r 2 values of between 0.92 and 0.99 for CO and values of between 0.77 and 0.87 for HC ŽStephens and Cadle, 1991.. Comparison of FEAT emission measurements with idle emission measurements has been less successful due to the dissimilarity in vehicle operating modes and speeds. However, quite a high degree of agreement was observed between FEAT results and a conventional system operated on a chassis dynamometer ŽHickman and McCrae, 1995.. Zhang et al. Ž1995. have reviewed the data collected from the use of the FEAT system to monitor vehicle exhaust emissions at 22 locations around the world. A strong relationship was observed between mean fleet emissions for CO and mean fleet emissions for HC due to common influencing factors such as differing levels of maintenance, average fleet age and imposition of compulsory emission standards. However, an individual vehicle that is a high emitter of CO is not necessarily a high emitter of HC. Quintile analy-
sis has been conducted on emissions data from four cities ŽGothenburg, Los Angeles, Melbourne and Leicester . and the higher CO and HC emissions for the highest emitting quintile of the newest vehicles compared with the lowest emitting quintile of the oldest vehicles indicates that in-use emission differences between well maintained and badly maintained vehicles is larger than the age-dependent deterioration of emissions.
2. Methodology and site characteristics 2.1. The FEAT system The FEAT system utilises standard spectroscopic principles and combustion equations to measure the CO and HC emissions from passing vehicles in under 1 s ŽBishop et al., 1989.. It consists of four major components: an infrared source, a detector, a computer and a video camera. A horizontal beam of infrared radiation is directed across a single lane of traffic approximately 30 cm above the road surface. As a vehicle breaks this infrared beam, measurements are taken of the infrared absorption both in front of the vehicle and as the vehicle exits the beam. The detector converts the incident infrared radiation to a voltage signal and the COrCO 2 and HCrCO2 ratios are measured to compensate for dilution effects which occur as the exhaust plume moves away from the vehicle. These ratios can be computed to the percentage exhaust composition of CO and HC. The system requires daily calibration with a certified gas mixture of CO, CO 2 and propane. The use of propane means that the hydrocarbon results are expressed in terms of ‘propane equivalents’. The video camera, which records 60 framesrs, is focused on the rear of vehicles as they pass the sensor and the image of each vehicle is recorded together with the date, time, and the percentage of CO, HC and CO 2 in the emissions. The video information is recorded on tape and if required, the vehicle can be identified according to model, make and age by comparison with the Drivers Vehicle Licensing Authority ŽDVLA. records.
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
93
Table 1 Characteristics of monitored sites Site
Name
Sampling date
Traffic flow rate Žvehiclesrh.
Vehicle speeds
No of valid hydrocarbon readings
A
Bounds Green Road, Haringey Dixons Bank, Middlesborough Abbey Road, Southwark Uppingham Road, Leicester
May 1994
511]780
8]32
June 1994
465]650
40]80
March and August 1995
127]376
4]50
October 1995
315]701
8]30
11 099 Ž85%. 3979 Ž53%. 7414 Ž74%. 12 223 Ž81%.
B C D
2.2. Monitoring site characteristics The four monitored sites were all single lane carriageways. The relevant important characteristics of each site including the ranges of traffic flow rates, the ranges of vehicle speeds and the number of valid HC readings are described in Table 1. The low number of valid HC readings at Site B is due to the unfortunate loss of hydrocarbon channel data for one complete day.
3. Discussion 3.1. Fleet emissions of hydrocarbons Analysis of the measured hydrocarbon emissions according to their percentage contribution
to the fleet and to the total fleet emissions shows that the majority of vehicles are low emitters and contribute little to total fleet emissions. Thus, at site A ŽHaringey., 69% of vehicles ŽHC emission range - 1000 ppm. contributed to only 24% of fleet emissions ŽFig. 1.. Sites B ŽMiddlesborough., C ŽSouthwark. and D ŽLeicester . produced similar results ŽTable 2.. Those vehicles in the emission category representing hydrocarbon concentrations of less than 1000 ppm in Middlesborough, Southwark and Leicester, accounted for 67, 75 and 65% of the fleets, respectively, but contributed only 5, 18 and 9%, respectively, of fleet emissions. Further analysis of data from site A shows that those vehicles emitting - 2000 ppm accounted for 89% of the fleet, but contributed only 50% to total fleet emissions. Further analyses of the data from sites B]D produced a similar trend to that
Table 2 Summary of hydrocarbon emission statistics Site
A B C D
Mean Žppm.
1000 1510 940 1552
Median Žppm.
590 420 350 450
Emission category - 1000 ppm
Emission category - 2000 ppm
Emission category ) 10 000 ppm
Vehicles Ž%.
Fleet emissions Ž%.
Vehicles Ž%.
Fleet emissions Ž%.
Vehicles Ž%.
Fleet emissions Ž%.
69 67 75 65
24 5 18 9
89 82 88 79
50 19 35 23
1 3 1 4
14 40 19 33
94
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
Fig. 1. Hydrocarbon emissions for site A ŽBounds Green Road, Haringey. expressed as percentage contribution to the fleet and the fleet emissions.
of site A with a large proportion of the fleet contributing comparatively little to fleet emissions ŽTable 2.. The converse situation is also true in that a small minority of highly polluting vehicles Žhydrocarbon emission concentrations ) 10 000 ppm. contribute a significant proportion to fleet emissions, with 3% of the fleet at site B producing 40% of fleet emissions. Site D produced a very similar result with 4% of the fleet producing 33% of total fleet emissions. However, sites A and C although agreeing with the general trend, only had 1% of the fleet in the emission concentration range ) 10 000 ppm which contributed 14 and 19% of the fleet emissions, respectively. This situation, although agreeing with the general pattern found for CO emissions ŽMuncaster et al., 1996., differs in that a greater proportion of the total fleet emissions are contributed by the highly polluting vehicles.
The mean emissions for sites A]D are 1000, 1510, 940 and 1552 ppm, respectively. The corresponding median values at each site are 590, 420, 350 and 450 ppm ŽTable 2.. The differences between the mean and median values reflect the very skewed nature of the emission distributions ŽFig. 1.. Calculation of the skewness values for the emission distributions shows site C to have the most skewed distribution with a skewness value of 2.52 although similar values were found at all sites Žsite A, 2.17; site B, 2.28; site D, 2.29.. Similar distributions and skewness values have been observed for CO at these sites ŽMuncaster et al., 1996.. The emission distributions are similar to those seen in other remote sensing studies carried out in the UK ŽVanke and Bidgood, 1992; Muncaster et al., 1994. and overseas ŽStedman et al., 1991b; Sjodin, 1994.. The mean emission level obtained for Leicester
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
is lower than the 2100 ppm value recorded by Zhang et al. Ž1995. for the same city but similar to the 1650 ppm found by Vanke and Bidgood Ž1992.. The 2100 ppm value would compare Leicester with Mexico City, Kathmandu and Bangkok in the hierarchical cluster analysis performed by Zhang et al. Ž1995.. The fleets in these cities are composed of older vehicles that are less well maintained compared to Leicester and not subject to emission control legislation or new emission standards. The results from this study are more realistic in that they place Leicester in direct comparison with cities where the fleets are newer, better maintained and are subject to emission control legislation. The mean emission levels for the two London sites are similar but are lower than the 1400 ppm found by Zhang et al. Ž1995. and the 1910 ppm obtained by Vanke and Bidgood Ž1992. . The London fleet exhibits similar hydrocarbon emission distributions to those of Edinburgh, Melbourne and Chicago according to Zhang et al. Ž1995. and the results obtained from both the Haringey and Southwark monitoring studies agree with this. In
95
making these comparisons it is important to recognise that remote sensing measurements obtained on roads in the same urban area may be influenced by different factors, such as traffic origin]destinations, which will affect the emission distribution. 3.2. Model year hydrocarbon emissions. The existence of a relationship between mean emissions and the age of the vehicle has been investigated by plotting mean hydrocarbon emissions against model year. The results, including error bars calculated using a 95% confidence interval, show a large degree of scatter with consistently low r 2 values at all sites. The Leicester fleet Žsite D. showed the highest association between model year and mean hydrocarbon emissions with a r 2 value of 0.26. The Haringey Žsite A., Middlesborough Žsite B. and Southwark Žsite C. fleets all gave r 2 values of 0.09, indicating that only 9% of the variation in mean emissions was explained by the model year of the vehicle. A strong influencing factor in this low correlation is
Fig. 2. Relationship between mean annual hydrocarbon emissions, with error bars calculated using a 95% confidence interval, and model years, 1985]1995, for Site D ŽUppingham Road, Leicester . Ž r 2 s 0.96..
96
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
the increased variability with vehicle age which is clearly illustrated by the large error bars on the older model year cars. The small sample sizes of these vehicles is also an important factor and a similar effect has been observed for CO ŽMuncaster, 1996.. Examination of the emissions data from site A shows that out of a total of 7017 valid vehicle measurements, only 912 measurements Ž13% of the fleet. were recorded for those vehicles produced before 1983. The equivalent emissions data statistics for sites B]D were 169 valid measurements Ž4% of the fleet., 350 valid measurements Ž8% of the fleet., 667 valid measurements Ž10% of the fleet., recorded, respectively, for those vehicles manufactured before 1983. For 1965, 1968 and 1969 there was only one vehicle measured for each model year at site A, with only two vehicles measured for years 1967 and 1970. For this reason, only those vehicles registered between 1983 and 1993 for sites A and B and 1985 and 1995 for sites C and D have been included in an additional investigation of the variation of hydrocarbon emissions with model year. The result is a considerable decrease in the size of the error bars and a scatter plot displaying the relationship between mean hydrocarbon levels and model year for site D is displayed in Fig. 2. The Leicester fleet together with that analysed for site C ŽSouthwark. showed the greatest association between the model year of a vehicle and mean pollutant emissions Ž r 2 values of 0.96 and 0.91, respectively.. The regression equations for the relationships found at these sites have similar large negative regression coefficient values of y172.77 and y190.63, respectively. This indicates that at both sites mean pollutant emissions change in a similar pattern such that a small increase in model year is accompanied by a large decrease in pollutant level. This relationship was also found for CO ŽMuncaster et al., 1996.. Sites A and B have similar regression coefficients of y92.52 and y68.8, respectively, so that in comparison to sites C and D, a larger increase in model year is required for the same decrease in pollutant level. This difference may be due to the additional catalyst equipped model years at sites C and D compared to sites A and B. These
additional catalyst-equipped model year vehicles increase the slope of the regression line by accentuating the contrast between the older non-catalyst-equipped vehicles and the more modern technology vehicles such that a small increase in model year is accompanied by a large decrease in pollutant level. The developed regression equations can be used to predict mean hydrocarbon emissions for different model years and these can be compared to the observed values. There is an over-prediction of mean emissions for model year 1993 at all sites which can be attributed to the introduction in January 1993 of EU Directive 91r441rEEC. This Directive introduced new emission limits which were most readily achieved by the fitting of threeway catalysts ŽTWC.. Therefore, a proportion of the hydrocarbon emissions not explained by the regression equations can be attributed to the introduction of catalyst technology to the UK fleet. Mean emissions for model year 1992 are also lower than predicted, and are explained by the early introduction of some catalyst-equipped vehicles ahead of Directive 91r441rEEC. In contrast, model years 1994 and 1995 at site D show an under-prediction of mean emissions which occurs because the steep decline in emissions from older non-catalyst-equipped vehicles to new catalyst-equipped vehicles is not sustainable. Due to all post-1992 model year vehicles having catalysts there is a plateauing of emissions at a very low base level and not a reduction, hence the under-prediction of emissions. However, it is important to note that EU Directive 91r441rEEC was only the latest of many vehicle emission regulations beginning with EU Directive 70r220rEEC Žinstituted to take measures against air pollution from vehicle engine emissions. for which the effect has been to reduce emissions from motor vehicles over time. 3.3. Model year contribution to fleet hydrocarbon emissions Although emissions generally increase with the age of the vehicle, older vehicles do not contribute significantly to fleet emissions due to the small number of old vehicles on the road. Analy-
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
sis of vehicle numbers and fleet emissions at site A shows that pre-1983 registered vehicles contribute only 18% of total fleet emissions. Site B fleet emissions demonstrate a similar scenario with pre-1983 registered vehicles Ž4% of the fleet. contributing only 5% to fleet emissions. Scrutiny of vehicle numbers and fleet emissions at sites C and D displays a similar pattern with only 12% and 16%, respectively, of total fleet emissions, contributed by pre-1983 manufactured vehicles. Because new vehicles are low emitters, it is not true that they do not contribute significantly to total fleet emissions. Analysis of vehicle numbers in relation to fleet emissions at site A shows that those vehicles registered during and after 1983 Ž80% of the fleet. contribute 82% of total fleet emissions. Site B emissions data reveals a similar pattern with vehicles registered during and after 1983 contributing 95% of total fleet emissions. Emissions data for sites C and D are 88% and 84%, respectively, of total fleet emissions being produced by vehicles registered during and after 1983. However, reduction of on road emissions by blanket targeting of this more modern age group would not necessarily produce the desired result. The vast majority of these vehicles are low emitters and it is the gross polluter component that
97
contributes significantly to total fleet emissions and therefore needs specific targeting. A gross polluter is defined as any vehicle that falls within the top 20% of emitters Žquintile 5. at a specific site when the model year is rank ordered. This is illustrated by an analysis of the emissions data from site A ŽFig. 3.. The gross polluter component of model year 1987, a mere 139 vehicles out of a fleet total of 7017 vehicles, contributed 7.7% of total fleet emissions. This pattern was repeated for those vehicles registered in 1988 with the gross polluter component Ž152 vehicles. contributing 8.1% of total fleet emissions. An analysis of the emissions data from site B also illustrates the impact high emitters have on fleet emissions with the gross polluter component of model year 1989 contributing 10.7% of total fleet emissions from just 2.3% of the fleet. A similar pattern was observed at sites C and D with the gross polluter component of model year 1988 at site C Ž1.9% of the fleet. contributing 9.5% of total fleet emissions. For those vehicles registered in 1985 at site D, the gross polluter component Ž1.5% of the fleet. contributed 8.9% of total fleet emissions. Therefore, fleet emissions could be significantly reduced if those vehicles in quintile 5 received regular emissions testingrservicing or were removed from the highway by the withdrawal of
Fig. 3. Hydrocarbon quintile emissions for site A ŽBounds Green Road, Haringey..
98
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99
their vehicle registration. An investigation of the site A data shows that if only those vehicles in quintile 5, registered between 1985 and 1991, were targeted then there is the potential to reduce fleet emissions at site A by 41%. The same analysis at sites B]D shows that fleet emissions could be reduced by 66, 51 and 48%, respectively. Rueff Ž1992. conducted a remote sensing survey to see if emission reduction was possible and concluded ‘that the repair of high-emitting vehicles identified via remote sensing constitutes a viable, cost-effective means of reducing certain types of vehicle emissions’. Similar work ŽBishop et al., 1993; Cadle et al., 1993. has shown that after repairs had been carried out emissions had decreased by more that 50% for the repaired vehicles. Older gross polluters are less important because there are so few vehicles in this category and therefore they have a small impact on total fleet emissions. An analysis of emissions from site A shows the minimal importance of older gross polluters ŽFig. 3.. If those vehicles registered in 1976 are examined, it can be shown, that only two vehicles were gross polluters, and they contributed only 0.2% to total fleet emissions. An analysis of emissions from the other sites shows a similar trend. The hydrocarbon quintile emissions distributions are similar in form to those of CO at these sites ŽMuncaster, 1996. and to those seen in other remote sensing surveys ŽStedman et al., 1991a,b; Guenther et al., 1994. with the greater contribution to fleet emissions coming from the gross polluters in quintile 5 with little contribution to fleet emissions from old vehicles.
4. Conclusions The results of this survey indicate that fleet emissions could be reduced markedly if only a small proportion of vehicles are repaired or removed from the road. One possible way of achieving this would be to employ remote sensing techniques to screen the in-use fleet for high emittingrgross polluting vehicles, which would then be subject to further test and repair. Remote
screening has the additional advantage of being performed without warning, thereby identifying vehicles that may otherwise have been tampered with before and after conventional inspection as well as those which attempt to avoid inspections completely. References Ashbaugh LL, Lawson DR, Bishop GA, Guenther PL, Stedman DH, Stephens RD, Groblicki PJ, Parikh JS, Johnson BJ, Huang SC. On-road remote sensing of carbon monoxide and hydrocarbon emissions during several vehicle operating conditions. In: AWMArEPA Conference on PM 10 Standards and Particulate Control. 1992; 9:1]10. Bishop GA, Starkey JR, Ihlenfeldt A, Williams WJ, Stedman DH. IR long path photometry, a remote sensing tool for automobile emissions. Anal Chem 1989;61:671a]677a. Bishop GA, Stedman DH, Peterson JE, Hosick TJ, Guenther PL. A cost-effectiveness study of carbon monoxide emissions reduction utilising remote sensing. J Air Waste Manage Assoc 1993;43:978]988. Cadle SH, Gorse RA, Lawson DR. Real-world vehicle emissions: a summary of the 3rd annual co-ordinating council-air pollution research advisory committee ŽCRC-APRAC.. J Air Waste Manage Assoc 1993;43:1084]1090. Guenther PL, Bishop GA, Peterson JE, Stedman DH. Emissions from 200, 000 vehicles: a remote sensing study. Sci Total Environ 1994;146r147:297]302. Hickman AJ, McCrae IS. Evaluation of a remote vehicle emission measurement system. Transport Research Laboratory Project Report No. 1052, 1995. Muncaster GM, Hamilton RH, Revitt DM, Stedman DH, Vanke J. Individual emissions from on-road vehicles. In: Proceedings of 27 th ISATA Conference on The Motor Vehicle and the Environment, 1994;29:267]274. Muncaster GM. Vehicle emissions and roadside air quality. Unpublished PhD thesis. London, UK: Middlesex University, 1996:282. Muncaster GM, Hamilton RH, Revitt DM. Remote sensing of carbon monoxide vehicle emissions. Sci Total Environ 1996;189r190:149]153. Rueff RM. The cost of reducing emissions from late-model high emitting vehicles detected via remote sensing. J Air Waste Manage Assoc 1992;42:921]925. Sjodin A. On-road emission performance of late-model TWC-cars as measured by remote-sensing. J Air Waste Manage Assoc 1994;44:397]404. Stedman DH, Bishop GA, Peterson JE, Guenther PL. On-road CO remote sensing in the Los Angeles Basin. Contract No. A932-189. Sacramento: CARB, 1991a. Stedman DH, Bishop GA, Peterson JE, Guenther PL, McVey IF, Beaton SP. On-road carbon monoxide and hydrocarbon remote sensing in the Chicago area. Report No. ILENRrRE-AQ 15, 1991b.
D.M. Re¨ itt et al. r The Science of the Total En¨ ironment 235 (1999) 91]99 Stedman DH, Bishop GA, Pitchford M. Evaluation of a remote sensor for mobile source CO emissions. Report No. EPAr600r4-90r032, 1991c. Stephens P, Cadle SH. Remote sensing of carbon monoxide emissons from road vehicles. J Air Waste Manage Assoc 1991;41:39]43.
99
Vanke J, Bidgood JFS. Remote sensing of vehicle emissions}principles and potential. Inst Mech Eng 1992: C387]C457. Zhang Y, Stedman DH, Bishop GA, Guenther PL, Beaton SP. Worldwide on-road vehicle exhaust emissions study by remote sensing. Environ Sci Technol 1995;29:2286]2294.