Transportation Research Part D 11 (2006) 242–249 www.elsevier.com/locate/trd

Roadside measurement and prediction of CO and PM2.5 dispersion from on-road vehicles in Hong Kong J.S. Wang a, T.L. Chan

a,*

, Z. Ning a, C.W. Leung a, C.S. Cheung a, W.T. Hung

b

a

b

Research Centre for Combustion and Pollution Control, Department of Mechanical Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

Abstract This study investigates the traffic induced gaseous and particle emissions dispersion characteristics from the typical urban roadside sites in Hong Kong. Concentrations of carbon monoxide, CO, fine particles, PM2.5 pollutants, and the traffic and ambient atmospheric conditions at three selected local urban road sites were simultaneously measured. A developed local general finite line source model (GFLSM) was used to predict the local roadside CO and fine particle concentrations. A high level of agreement found between the measured and calculated CO and PM2.5 data. Generally, the roadside concentrations of gaseous and PM2.5 pollutants decrease with the distance away from the road and the exposure to both gaseous and particle pollutants in the vicinity of the selected urban road sites is interrelated to on-road vehicle emissions. It has also demonstrated that the developed local general finite line source model has the capability of reasonably predicting the characteristics of gaseous and particle pollutant dispersion from on-road vehicles for the local urban air quality. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Finite line source dispersion model; Particulate matter, PM2.5; Carbon monoxide; On-road vehicles; Field measurement

1. Introduction Motor vehicle emissions are the major source of air pollution problem in most urban cities. Many models have been developed and applied to simulate the line source emission dispersion in the rural open highway, such as the General Motors (GM) line source model (Chock, 1978), California line (CALINE) source models (Benson, 1979) and Highway (HIWAY) air pollution models (Zimmerman and Thompson, 1975). Luhar and Patil (1989) presented a general finite line source model (GFLSM) for all wind directions based on the Gaussian diffusion equations. The prediction performance of GFLSM was compared with GM, CALINE-3 and HIWAY-2 models and a reasonable accuracy for the Indian traffic conditions was shown (Sharma and Khare, 2001; Nagendra and Khare, 2002). In the present study, a local GFLSM model has been developed for simulating the CO and fine particle dispersion from the selected urban road sites in Hong Kong. *

Corresponding author. Tel.: +852 2766 6656; fax: +852 2365 4703. E-mail address: [email protected] (T.L. Chan).

1361-9209/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.trd.2006.04.002

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

243

2. The general finite line source model 2.1. GFLSM model equation for gaseous pollutant dispersion For the prediction of traffic-induced gaseous pollutants, the developed line source model is based on the classical GFLSM (Luhar and Patil, 1989) and the GM line source model (Chock, 1978). Luhar and Patil (1989) first developed the GFLSM by using the coordinate transformation between the wind and the line source coordinate systems as: " ! !# Q ðz  he Þ2 ðz þ he Þ2 C ¼ pffiffiffiffiffiffi exp  þ exp  2r2z 2r2z 2 2prz ue " ! !# sin hðL=2  yÞ  x cos h sin hðL=2 þ yÞ þ x cos h pffiffiffi pffiffiffi  erf þ erf ð1Þ 2ry 2r y where C is the concentration (mg m3 or ppm); Q, the line source strength per unit length (mg m1 s1 or m3 m1 s1); ry and rz are the horizontal and vertical dispersion parameters (m), respectively; ue = u sin h + u0 where u, the mean ambient wind speed (m s1) at a specific source height; u0, the wind speed correction due to the traffic wake. A constant numerical value, u0 = 0.2 ms1 is applied for the calculation; h is the angle between the roadway and wind direction; he = h + hp is the effective source height, where h, the line source height (m) and hp is the plume rise (m); L, the line source length (m). 2.2. GFLSM model equation for particle pollutant dispersion For the prediction of traffic-induced particle pollutants, the particle settling velocity needs to be considered but the reflection at the earth’s surface is ignored. This is because gravity has a greater effect on the particles/ particulates movement than the gaseous pollutants one (Luhar and Patil, 1989). The basic equations of the developed model for particle emissions prediction is: 8 h 2  i2 93 V ti x > > N = < z  h  X e ð uþu0 ÞðAþB sin hÞ wi Q 6 7 pffiffiffiffiffiffi C¼ 4exp  5 2 > > 2r 2 2p r u ; : z z e i¼1 "

(

sin hðL=2  yÞ  x cos h pffiffiffi  erf 2ry

)

(

sin hðL=2 þ yÞ þ x cos h pffiffiffi þ erf 2ry

)# ð2Þ

where N is the number of particle size classes; V ti , the settling velocity corresponding to the average particle size of the ith class; wi, the weight fraction of particulates in the ith size class; the coefficients of A and B are stability dependent which are available from the GM line source model (Chock, 1978). Assuming the particulate matters to be spherical, the settling velocity, Vt of the particle can be determined as: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 4Dp gðqp  qa Þ ð3Þ Vt ¼ 3C D qa where Dp is the diameter of particle; g, the gravitational acceleration; qp, the density of particle; qa, the density of the ambient air fluid; CD, the drag coefficient of particle; g, the gravitational acceleration. With the simplified formula, for example for the laminar flow, CD can be obtained from the Stokes law. The Reynolds number of particle can be determined by using the Archimedes number: Ar ¼

D3p gqa ðq  qa Þ la

ð4Þ

244

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

When the Archimedes number, Ar < 1.83, the Reynolds number of particle, Rep can be calculated by Rep ¼

Ar 18

ð5Þ

If the diameter of particle, Dp 6 2.5 lm, then the settling velocity of particle, V ti 6 0:0002 m s1, so the term V ti x in Eq. (2) can be neglected if the downwind distance, x is also small. In the present study, the ðuþu0 ÞðAþB sin hÞ prediction dispersion of PM2.5 (i.e., Dp 6 2.5 lm) from the roadside can be simplified from Eq. (2) as: " ( )# N X wi Q ðz  he Þ2 pffiffiffiffiffiffi C¼ exp  2r2z i¼1 2 2prz ue " ( ) ( )# sin hðL=2  yÞ  x cos h sin hðL=2 þ yÞ þ x cos h pffiffiffi pffiffiffi þ erf ð6Þ  erf 2ry 2ry 2.3. Dispersion parameters The dispersion parameters are the function of the Monin–Obukhov length, the friction velocity and the mixing height (instead of the discrete Pasquill classes). These parameters are computed using the meteorological pre-processing model. The lateral and vertical dispersion parameters (ry and rz) can be written as: r2y ¼ r2ya þ r2y0 ; r2z ¼ r2za þ r2z0

ð7Þ

where the subscripts of a and 0 refer to atmospheric turbulence and traffic-originated turbulence, respectively. The atmospheric turbulence is evaluated at the effective distance from the line source and at the effective source height. The Pasquill parameters are determined from Pasquill, 1961. However, the vertical dispersion parameter for atmospheric turbulence can be determined from the GM line source model. The dispersions due to the traffic-induced turbulence are: rz0 ¼ 3:57  0:53U c ry0 ¼ 2rz0 U c ¼ 1:85u0:164 cos2 h

ð8Þ

2.4. Determination of Monin–Obukhov length Monin–Obukhov length, L, can be expressed as: L¼

T 2 u2 jgh

ð9Þ

where T2 is air temperature at the height of 2 m; j is von Ka`rma`n constant, 0.41; u*, the friction velocity; and h*, a temperature scale for turbulent heat transfer. The u* and h* are determined from van Ulden (1978), and are then iterated when the single wind speed at z1 and the temperature difference at z3 and z2 are measured at the selected local urban roadside site. 3. Traffic-induced air pollutants from on-road vehicles using the developed local GFLSM The roadside measurements of traffic-induced CO and PM2.5 concentration data were collected continuously at several selected sampling points for a certain sampling period. The local meteorological data, such as the ambient temperature, wind speed and direction at the roadside sites, were collected continuously as the input parameters for the developed local GFLSM. Traffic information, including the average vehicle speed and the traffic flow rate from different vehicle categories during the sampling time interval, was also recorded.

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

245

3.1. Urban roadside experimental setup The roadside measurements from the traffic-induced emissions were carried out at three selected local urban road sites, namely Lion Rock Tunnel (Kowloon entrance), New Territories, Site 1; Waterloo Road and Durham Road crossroad area, Kowloon Tong, Site 2; Tai Po Road (near the parking lot area of Shatin Jockey Club), New Territories, Site 3. The selected roads are relatively open with no adjacent high buildings and the traffic flow rate is relatively high and steady. The typical experimental set up and location of pollutant sampling points at each selected local urban road site is shown in Fig. 1. No. 1 was located 1 m from the edge of the road. The relative distances between the sampling points from No. 1 to No. 2, No. 2 to No. 3 and No. 3 to No. 4 at each site are shown in Table 1. All the sampling points approximate the breathing level of human body; namely 1.5 m above the ground level. The concentrations of PM2.5 at locations 1, 2 and 4 were monitored using aerosol monitors (Model 8520, TSI DustTrak, USA) as shown in Fig. 1. Corresponding concentrations of CO were collected using sampling bags at all sites and analyzed using the gas filter correction CO analyzer (Model 48, Thermo Environmental Instruments Inc., USA). Sampling point, No. 4 was taken as the ambient background conditions. To investigate the traffic-induced emissions dispersion characteristics for different local urban road traffic conditions, three typical sampling periods such as morning sampling period (8:00 to 10:00); noon sampling period (11:00 to 13:00) and afternoon sampling period (15:00 to 17:00) were designated. For each designated sampling period, at least three tests were carried out to collect the roadside CO and PM2.5 pollutant concentration data. The sampling time was 5 min for each test. Location 1 represents the local ambient weather conditions were monitored using the combined wind velocity and direction sensors, temperature and humidity sensors, and data acquisition device (Model DNA022, DMA570, and BSA020.E, LSI Spa, Italy). Meanwhile, the air temperature and wind speed at two vertical locations were measured to determine the Monin–Obukhov length using the thermo-anemometer (TSI VeloCalcu Plus Model 8386) and was used to describe the atmospheric stability conditions.

Road

No.1 CO, PM2.5

No.2 CO, PM2.5

No.3 CO

No.4 CO, PM2.5

Fig. 1. Distribution of the sampling points (Nos. 1–4) at each selected local urban roadside site. Table 1 Distances between the sampling points at the selected local urban roadside sites Roadside site

Site 1 (Lion Rock Tunnel, Kowloon entrance)

Site 2 (Waterloo road)

Site 3 (Tai Po road)

Sampling points

Distance (m)

Distance (m)

Distance (m)

No.1–No.2 No.2–No.3 No.3–No.4

10.0 10.0 10.0

56.0 46.0 53.2

16.5 27.8 29.4

246

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

Table 2 Traffic and weather conditions at the selected local urban roadside sites Roadside site

Average traffic flow density (vehicles h1)

Average traffic speed (km h1)

Main wind direction

Wind speed (m s1)

Ambient temperature (°C)

1 2 3

2337 6790 5724

20 34 62

Perpendicular to road Parallel to road Parallel to road

1.0 to 1.8 0.8 to 1.9 0.5 to 2.2

21.9 23.4 20.6

Table 2 shows the average local traffic and weather conditions at each selected local urban roadside site. The average traffic flow density at Site 2 was the highest and at Site 1 was the lowest. The traffic volumes of the road were fairly high for these three selected local urban road sites. 3.2. Mobile source emission rate and emission factor Emission rates depend on the volume of traffic, its composition, and the operating modes of the vehicles. The line source strength is estimated using the polynomials obtained from emission rate of vehicles with running vehicle speed for seven typical local vehicle types in Hong Kong (i.e., bus and light bus, light duty goods, middle duty goods and heavy duty goods, taxi and private cars). The general form of the polynomials is: ri ¼

N X

an V n

ð10Þ

n¼0

where i is the vehicle type (i = 1,2,..,7 represents the bus, light bus, light duty goods, middle duty goods and heavy duty goods, taxi and private cars, respectively); r, the emission rate in g km1veh1; an are the modeling coefficients; V, the vehicle driving speed in km h1. The line source strength in mg m1 s1 can be given by: Q¼

7 X

2:7778  104 ri trvi

ð11Þ

i¼1

where trvi is the traffic flow density per hour in veh h1. The vehicle emission factors of CO were based on the aggregate on-road vehicles as a function of instantaneous speed profiles in Hong Kong from Chan et al. (2004), Chan and Ning (2005), Ning and Chan (2006). While the PM2.5 emission factors were based on the established correlation equations from Ning et al. (2004) and the emission factor database for UK of European standard vehicles (Inventory, 2003). 4. Results 4.1. Comparison of the measured and predicted emission concentrations The averaged concentrations of CO and PM2.5 for each sampling point and site during the sampling period were compared with the predicted concentrations using the developed local GFLSM as shown in Figs. 2 and 3. A general decreasing trend of the measured CO and PM2.5 concentrations with increasing distance from the near roadside at the selected local urban road sites was obtained. The measured CO and PM2.5 concentrations decrease 13% and 11% with increasing distance, 30 m from the near roadside, respectively at the Site 1. The measured CO and PM2.5 concentrations decrease 36% and 18% with increasing distance, 155 m from the near roadside, respectively at the Site 2. The measured CO and PM2.5 concentrations decrease 21% and 24% with increasing distance, 75 m from the near roadside, respectively at the Site 3. Similar decreasing trend of the predicted CO and PM2.5 concentrations data with increasing of distance from the road was also calculated from these three selected local urban sites based on the developed local GFLSM. The results demonstrate that the average traffic flow speed, the combination of vehicle fleets, geometrical and meteorological conditions play a significant role on the dispersion characteristics of the traffic-induced gaseous and particle emissions at the urban roadside site. The comparison of the measured and predicted gaseous emissions of CO and particle

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

247

CO concentration (ppb)

2000 measured data in Site 1 predicted data in Site 1 measured data in Site 2 predicted data in Site 2 measured data in Site 3 predicted data in Site 3

1500

1000

500 0

40

80

120

160

Distance from the road (m) Fig. 2. Comparison of measured and predicted CO concentration at the selected local urban roadside sites.

100 measured data in Site 1 predicted data in Site 1 measured data in Site 2 predicted data in Site 2 measured data in Site 3 predicted data in Site 3

PM2.5 concentration ( μgm-3)

90 80 70 60 50 40 30 20 0

40

80

120

160

Distance from the road (m) Fig. 3. Comparison of measured and predicted PM2.5 concentration at the selected local urban roadside sites.

emissions of PM2.5 show that the developed local GFLSM has a reasonable prediction performance for both gaseous and particle emissions with an average prediction error within 10% for CO and an average prediction error within 13% for PM2.5 at these three selected local urban roadside sites. The developed local GFLSM model overpredicts the concentrations of CO and PM2.5 at Site 1 within 6% and 10%, respectively, but underpredicts the concentration of CO within 13% and 8%, and the concentration of PM2.5 within 16% and 18% at Site 2 and Site 3, respectively. The prediction performance of the developed local GFLSM depends on a lot of factors, such as the effect of ambient wind speeds and directions, the accuracy of the estimation of trafficinduced emissions etc. In the present study, the average ambient wind speed was 1.5 m s1 at Site 1, 0.5 m s1 at Site 2 and 0.3 m s1 at Site 3. The ambient air conditions have direct effect on the prediction performance of this developed model. The model tends to overpredict the gaseous and particle pollutants when ambient wind speed is high and underpredict the gaseous and particle pollutants when the ambient wind speed is low. 4.2. Analysis of the GFLSM evaluation The average statistical analysis parameters for all the tests at different selected local urban roadside sites are shown in Table 3. The values of index of agreement range from 0.68 to 0.92 for CO concentrations and from

248

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

Table 3 Average statistical parameters for model evaluation at the selected local urban roadside sites Emissions (Unit)

Summary Observed Predicted Observed Predicted

measures mean mean deviation deviation

Site 1 Lion Rock Tunnel (Kowloon entrance)

Site 2 Waterloo road

Site 3 Tai Po road

CO (ppb)

PM2.5 (lg m3)

CO (ppb)

PM2.5 (lg m3)

CO (ppb)

PM2.5 (lg m3)

762.1 814.4 45.0 25.4

88.0661 90.2156 4.3323 3.0528

1362.9 1203.8 273.7 58.5

50.4614 44.0379 4.7229 4.6322

884.2 846.3 84.4 73.1

55.4177 50.0092 6.1933 6.2140

Linear regression Intercept Slope Correlation coefficient

0.4320 0.5051 0.9309

33.1293 0.6546 0.9032

0.8984 0.2268 0.9901

6.7308 1.0163 0.9625

0.0441 0.9213 0.9497

2.6801 0.8585 0.8472

Index of agreement

0.6791

0.5997

0.9189

0.7128

0.8400

0.7524

0.60 to 0.75 for PM2.5 concentrations. All of these values correspond to a fairly good agreement between the predicted and measured values at the roadside sites. The linear regression analysis for the predicted and measured concentrations shows high coefficient values ranging from 0.93 to 0.99 for CO concentrations and from 0.85 to 0.96 for PM2.5 concentrations which indicates that the model predicts the shape of the CO and PM2.5 concentration profiles correctly. The developed local GFLSM has provided a reliable and convenient tool to evaluate the urban air quality at roadside in Hong Kong. The present study has established a research methodology in estimating the trafficinduced emissions characteristics for local urban air quality at roadside for carrying out similar investigations at different sites in full campaigns. 5. Conclusions The traffic induced gaseous and particle emissions dispersion characteristics from the typical urban roadside sites in Hong Kong have been investigated. The roadside concentrations of carbon monoxide, CO and fine particles, PM2.5 pollutants, and the traffic and ambient atmospheric conditions at three selected local urban road sites were simultaneously measured. The developed local general finite line source model (GFLSM) was used to predict the local roadside CO and fine particle concentrations. The atmospheric stability condition was determined based on the Monin–Obukhov length. A good agreement has been obtained based on their measured and calculated CO and PM2.5 data. Generally, the roadside concentrations of gaseous and PM2.5 pollutants decrease along the distance away from the road and the exposure of both gaseous and particle pollutants in the vicinity of the selected urban road sites is interrelated to the on-road vehicle emissions in Hong Kong. It has also demonstrated that the developed local GFLSM model has good capability in predicting the characteristics of gaseous and particle pollutant dispersion from on-road vehicles for the local urban air quality. Acknowledgements This work was supported by Grants from the Research Grants Council of the Hong Kong Special Administrative Region, China (RGC Project No. PolyU 5154/01E) and the Central Research Grants of The Hong Kong Polytechnic University (Project No. B-Q497). References Benson, P.E., 1979. CALINE-3. A versatile dispersion model for predicting air pollutant levels near highway and arterial roads. Final Report, FHWA/CA/TL.-79/23, California Department of Transportation, Sacramento.

J.S. Wang et al. / Transportation Research Part D 11 (2006) 242–249

249

Chan, T.L., Ning, Z., 2005. On-road remote sensing of diesel vehicle emissions measurement and emission factors estimation in Hong Kong. Atmospheric Environment 39, 6843–6856. Chan, T.L., Ning, Z., Leung, C.W., Cheung, C.S., Hung, W.T., Dong, G., 2004. On-road remote sensing of petrol vehicle emissions measurement and emission factors estimation in Hong Kong. Atmospheric Environment 38, 2055–2066. Chock, D.P., 1978. A simple line source model for dispersion near roadways. Atmospheric Environment 12, 823–829. Luhar, A.K., Patil, R.S., 1989. A general finite line source model for vehicular pollutant prediction. Atmospheric Environment 23, 555– 562. Nagendra, S.M.S., Khare, M., 2002. Line source modeling. Atmospheric Environment 36, 2083–2098. National Atmospheric Emission Inventory 2003. Exhaust Emission Factors 2003: Database of Emission Factors, National Atmospheric Emission Inventory of the United Kingdom, Available from: <(http://www.naei.org.uk/emissions/index.php)>. Ning, Z., Chan, T.L., 2006. On-road remote sensing of liquefied petroleum gas (LPG) vehicle emissions measurement and emission factors estimation in Hong Kong, working paper. Ning, Z., Chan, T.L., Wang, J.S., Cheung, C.S., Leung, C.W., Hung W.T., 2004. Relationship between ultrafine particle, fine particle, coarse particle and hydrocarbon (HC) emission factors from the selected local representative in-use vehicles in Hong Kong. Motor Vehicle Emission Control Workshop 2004 (MoVE2004), Hong Kong. Pasquill, F., 1961. The estimation of the dispersion of windborne material. Meteorological Magazine 90, 33–49. Sharma, P., Khare, M., 2001. Modelling of vehicular exhausts – a review. Transportation Research Part D 6, 179–198. van Ulden, A.P., 1978. Simple estimates for vertical diffusion from sources near the ground. Atmospheric Environment 12, 2125–2129. Zimmerman, J.P., Thompson, R.S., 1975. User’s Guide for HIWAY. A Highway Air Pollution Model, EPA-650/4-74-008.