- Email: [email protected]
Evaluation of GRAL for the pollutant dispersion from a city street tunnel portal at depressed level
Environmental Modelling & Software 20 (2005) 499–504 www.elsevier.com/locate/envsoft
Short communication
Evaluation of GRAL for the pollutant dispersion from a city street tunnel portal at depressed level Dietmar Oettl*, Peter Sturm, Raimund Almbauer Institute for Internal Combustion Engines and Thermodynamics, Graz University of Technology, Inffeldgasse 21, 8010 Graz, Austria Received 31 May 2004; accepted 23 June 2004
Abstract A new data set for evaluation of atmospheric dispersion models, designed to treat road tunnel portals, is presented. Further, the performance of GRAL 3.5 (Graz Lagrangian model Version 3.5) using this data is discussed. One main drawback of the model in its current version is that two empirical parameters have to be adjusted for each tunnel site separately in dependence on average traffic volume and construction of the tunnel portal. As the model has been tested for five different tunnel sites by now, which differ significantly in traffic volume, terrain, and construction, model users should be able to choose reasonable values for those empirical parameters for other tunnel sites as well. GRAL has not been tested against field data, where the tunnel air is colder than ambient air. Here, further research is necessary. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: GRAL; Tunnel portal; Dispersion; Tunnel jet; Lagrangian model
Software availability GRAL: Free of charge from the authors for the operating systems Windows XP and Windows 2000 Professional.
1. Introduction Often difficulties occur regarding the optimum location of tunnel portals, as in many cases the whole vehicle emissions inside the tunnel are concentrated there. Dispersion models have to be applied to evaluate air quality standards at nearby residential areas. Dispersion of pollutants from tunnel portals requires new approaches in modelling as it differs significantly compared to dispersion of pollutants from other sources * Corresponding author. Tel.: C43 316 873 8081; fax: C43 316 873 8080. E-mail address: [email protected] (D. Oettl). 1364-8152/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.envsoft.2004.06.001
(e.g. line sources). The polluted air streaming out of the tunnel portal (hereafter called tunnel jet) is strongly influenced by the ambient wind in a way, that it is being forced to change the direction according to it. As was found by Oettl et al. (2002), the ambient wind direction fluctuations force the tunnel jet also to change its position, and in this way forming a very effective dispersion process. For convenience, this dispersion process will be abbreviated by the acronym ADAPT (Additional diffusion by the influence of a fluctuating ambient wind field on the position of the tunnel jet). In addition, buoyancy may also be an important factor for the dispersion of pollutants from tunnel portals, due to temperature differences between the tunnel air and the ambient air (Okamoto et al., 2001). In order to have a well suited dispersion model based on the current understanding of the particular processes described above, a novel method has been developed in the frame of a research project funded by the Austrian science fund (No. 14075-TEC). The new approach is
500
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
described in Oettl et al. (2002) and is incorporated as an individual module in the model GRAL (Graz Lagrangian Model, Oettl et al., 2003a). The module has been tested against experimental data from four different tunnel portals, namely the Enrei, Hitachi, Ninomiya tunnel in Japan (Oettl et al., 2003b), and the Ehrentalerbergtunnel in Austria (Oettl et al., 2002). The Japanese tunnels are all located in complex terrain, while the Austrian tunnel is surrounded by rather flat topography. Meteorological conditions observed during the experiments showed a wide variation of wind speeds (0.6– 6.2 m s1), atmospheric stabilities (stable–unstable), and wind directions, where it was able to evaluate the model for a wide range of angles between tunnel jets and ambient winds including head winds. In the year 2000 another research project was launched by the Austrian ministry for traffic, innovation and technology. The project aimed at investigating the pollutant dispersion in the proximity of a city tunnel in Vienna (Kaisermuehlentunnel), where the portal is situated 5 m below the surroundings. This is a particular difference in construction compared to all the other tunnel sites studied by now, which has an effect on the dispersion. Also, the traffic volume found in the Kaisermuehlentunnel is much higher than that for the other tunnels mentioned above. In the frame of this project, four different models (ADMS, LASAT, MUMO, and GRAL) were compared (Puxbaum et al., 2003). On the basis of this comparison, it was decided to recommend GRAL for dispersion modelling from road tunnel emissions in Austria in a new national guideline in elaboration.
2. Experimental set-up at the Kaisermuehlentunnel The Kaisermuehlentunnel in Vienna (Austria) has a length of 2150 m and two bores for each direction with six lanes in total. The pollution dispersion was studied at the south-east tunnel portal, where there exist two additional lanes to exit or enter the highway (Fig. 1). Continuous air quality observations were performed at five locations 2.5 m above ground level. Among the various chemical species recorded, NOx was found to be most related with the traffic on the highway and the tunnel jet. Another sampling point was set-up inside the tunnel to determine the emissions from the tunnel portal. The necessary volume flux was derived by recording the flow velocity in the tunnel using a cup anemometer. The method was validated with additional tracer tests by means of N2O releases inside the tunnel in five cases. Both methods agreed within G10%. The meteorological data used for modelling are based on wind observations with a cup anemometer on a 10 m mast at site M1 (Fig. 1). The experimental investigation lasted over a period of 10 months, where data were recorded on a half-hourly basis. As considerable background concentrations for NOx were expected, two distinct meteorological conditions were considered for the model evaluation: first, wind directions between 230 and 275 (979 cases) and second, wind directions between 95 and 125 (826 cases). In the first case the background concentration of NOx could be determined by use of sampling point M1, and in case of easterly winds, the average of sampling points M2–M5 was taken as background concentration.
Fig. 1. Orthophoto of south-east portal of the Kaisermuehlentunnel in Vienna (Austria). M1–M5 denote the sampling points for NOx. Meteorological data were recorded on a 10 m mast at site M1 using a cup anemometer.
501
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
Altogether 1805 cases were selected from the database for the simulations. The average wind speed found during the selected meteorological situations was 3.1 m s1. The maximum and the minimum wind speeds were 10.2 m s1 and 0.1 m s1, respectively. Temperature differences between the tunnel air and the ambient air were observed between 9.1 K and C14.2 K, and exit velocities of the tunnel jet ranged between 0.9 m s1 and 6.5 m s1. The mean NOxemission at the portal was 3.9 kg h1. In order to accurately model the NOx-concentrations, it is necessary to take into account the NOx-emissions resulting from the lanes out and into the tunnel. For the lanes out of the tunnel the corresponding NOx-emissions were determined by the NOx-emission at the portal divided by the length of the tunnel. As traffic data were not available on an half-hourly basis, NOx-emissions for the lanes into the tunnel were roughly estimated by assuming the same amount of traffic as out of the tunnel. This assumption clearly increases the uncertainty regarding the modelled mean half-hourly concentrations, while one can expect that it is a good estimation for the average concentrations over the whole period. The background concentrations for NOx were 35 mg m3 for easterly wind directions and 25 mg m3 for westerly winds.
3. Model set-up The dispersion model GRAL 3.5, respectively the tunnel module part of it, is described in detail in Oettl et al. (2002). It was already suggested in Oettl et al. (2002) to adjust some empirical parameters in the model to account for conditions, which differ in terms of construction and traffic volume from those, where the model has been validated by now. For the Kaisermuehlentunnel, the vertical dispersion has not been changed, while the ADAPT has been modified. The horizontal position of the tunnel jet is modelled by simulating the along and cross wind component of the jet, which depends on the ambient wind.
b, g the empirical constants; t the dispersion time [s]; UnS the cross wind component of the tunnel jet [m s1]; and UnA the ambient wind component perpendicular to the tunnel jet [m s1]. In case of the Ehrentalerbergtunnel, which has a rather low traffic volume of only a few hundred vehicles per hour, the parameters a and g were set equal to 1 (Oettl et al., 2002), while in case of the three Japanese tunnel sites, which have a traffic volume of about 1000–1400 vehicles per hour, a and g were set equal to 0.50 (Oettl et al., 2003b). The Kaisermuehlentunnel has a traffic volume of about 3740 vehicles per hour and furthermore, the tunnel portal is situated some 5 m below the surface. Thus, it is not surprising, that the ADAPT can be expected to be much lower compared to the other tunnel sites mentioned before. Therefore, a value of 0.003 (almost no change of the tunnel jet orientation) has been used for a and g in the simulations, except for head winds (south-easterly winds), where a value of 0.60 was chosen in order to obtain reasonable results. The motivation to take a higher value in case of head winds came from the consideration, that the ambient wind might not be influenced by the construction as it is the case for all other wind directions. It is apparent, that the values for a and g can vary over a wide range. Thus, it is quite difficult for users to choose those parameters properly. In order to make it easier for users to specify the values, a so-called stiffness parameter has been introduced, which is related with the parameters a and g as shown in Fig. 2. Table 1 provides guiding values for the stiffness parameter in dependence on traffic volume and construction of the tunnel portal. The stiffness parameter can take values between 0 and 100% depending on the traffic volume and construction of the tunnel portal. In order to get an estimate for the roughness length, sonic anemometer data were available to determine the
1 0.9
ð1Þ
KZað1CtÞ
ð2Þ
dUnS 1 2 Z bUnA 2 dt
ð3Þ
bZgð1CtÞ
ð4Þ
0.8
Parameters α and γ
dUp v2 Up ZK 2 dt vy
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
20
40
60
80
100
Stiffness parameter [%]
where Up is the along wind component of the tunnel jet [m s1]; K the turbulent exchange coefficient [m2 s1]; a,
Fig. 2. Relation between the stiffness parameter and the empirical parameters a and g.
502
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
Table 1 Guiding values for the stiffness parameter for a tunnel jet Tunnel
Stiffness parameter
Wind direction
Construction
Traffic [veh./h]
Terrain
Enrei
20–50 70
Head wind Other directions
At grade
w1300
Complex
Hitachi
20–50 70
Head wind Other directions
At grade
w1000
Complex
Ninomiya
20–50 70
Head wind Other directions
At grade
w1250
Complex
Ehrentalerberg
0 0–20
Head wind Other directions
5 m below the surface
w500
Flat
Kaisermuehlen
0–30 90
Head wind Other directions
5 m below the surface
w3800
Flat
friction velocity and subsequently the roughness length using u) z uðzÞZ ln ; ð5Þ z0 k where u(z) is the ambient wind speed depending on the height above ground in m s1, u* is the friction velocity as observed by the sonic anemometer in m s1, k is the Ka´rma´n constant taken to be 0.35, z is the height above ground level in m, and z0 is the roughness length in m. Eq. (5) is valid for neutral conditions only. A value of 0.17 G 0.08 m was obtained from the sonic anemometer recordings for the roughness length. Simulations were performed with a 5 m ! 5 m horizontal grid, and a vertical spacing of 1 m.
4. Results As most of the dispersion models, the tunnel module in GRAL 3.5 is based on classified average turbulence conditions. This means that relatively large uncertainty arises concerning modelled concentrations compared with observed ones paired in space and time, while average concentrations at fixed receptors will be more accurately modelled. Fig. 3 depicts the modelled average NOx-concentrations for the westerly winds and Fig. 4 shows the same but for easterly winds. The mean concentrations observed for the westerly winds were 38 mg m3, 23 mg m3, 60 mg m3, and 147 mg m3 for monitoring stations M2, M3, M4 and M5, respectively. It is
Fig. 3. Mean simulated concentrations for NOx for westerly winds in mg m3.
503
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
Fig. 4. Mean simulated concentrations for NOx for easterly winds in mg m3.
elled. Fig. 5 depicts the cumulative frequency distribution of the observed and modelled NOx-concentrations at site M5 for the westerly winds. As the model is based on average turbulence conditions, one cannot expect that it predicts the highest or lowest percentiles of concentration distributions perfectly. It is therefore not surprising that an underestimation of higher percentiles and an overestimation of lower percentiles was found, while the median value was modelled reasonably. Fig. 6 shows the same as Fig. 5 but for the monitoring site M1 for easterly winds. At this site a clear underestimation of the median concentration is obtained with GRAL 3.5. This might be regarded as an improper treatment of head wind situations, disturbances of the tunnel jet due to the portal itself under such conditions, or to the sharp variation of concentrations found near M1.
1.0 0.9
Cumulative frequency
remarkable, that M2 shows higher concentrations than M3 although for westerly winds it was expected that M2 should be less effected by the tunnel and road emissions than M3. Most probably the reason for the somewhat higher concentration at M2 is due to traffic emissions on the smaller streets adjacent to M2. These are not included in the background concentrations observed at M1. There is also a smaller street passing by the stations M3–M5, but it has lower traffic than all the small streets together crossing close to M2. Modelled NOx-concentrations with GRAL 3.5 give 8 mg m3, 21 mg m3, 60 mg m3, and 128 mg m3 at sites M2–M5. A rather good agreement was obtained for all monitoring sites except for M2, where the model suggested a substantial lower concentration than observed due to the reason mentioned above. In case of easterly winds, the modelled concentration at M1 was 110 mg m3 compared to observed 120 mg m3. The largest deviation between the average modelled and observed concentration was therefore about 13%, except at M2. Simulations were also performed for the tunnel jet alone without including the lanes into and out of the tunnel. In this way the contribution of the tunnel jet to the observed concentrations could be estimated. At sites M1 and M5 the contribution of the tunnel jet to the recorded mean concentrations (without background concentration) was about 52% and 60%, while at the monitoring site M3 the contribution was only about 10%. This circumstance is one of the main drawbacks of the data set, as it is not possible to evaluate the model performance for the tunnel module alone, but the roads have also to be considered. For practical purposes, it is also of interest how high percentiles of concentration distributions can be mod-
0.8
Observed GRAL3.5
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0
200
400
600
800
1000
Concentration [µg/m3] Fig. 5. Cumulative frequency distribution for the observed and modelled NOx-concentrations at site M5 for westerly winds.
504
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504 1.0
Cumulative frequency
0.9 0.8 Observed GRAL3.5
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0
100
200
300
400
Concentration
500
600
700
[µg/m3]
Fig. 6. Cumulative frequency distribution for the observed and modelled NOx-concentrations at site M1 for easterly winds.
In our opinion, a statistical evaluation of the model performance when comparing half-hourly observed and modelled concentrations at each site is not very helpful, as has already been outlined, the model is not designed for such applications. Furthermore, in this study, there exists relatively large uncertainty regarding modelled concentrations on a half-hourly basis due to the lack of knowledge of the actual traffic emissions on the lanes into the tunnel. Finally, a sensitivity analysis was made in setting the empirical coefficients a and g equal to 0.012, a value which is somewhat between the ones chosen for head winds and all the other wind directions before. The resultant modelled concentrations were 87 mg m3, 135 mg m3, 106 mg m3, and 52 mg m3 at the monitoring sites M3–M5 and M1. While concentrations at M3 and M4 were substantially overestimated by more than a factor of two, concentrations at M5 and M1 were underestimated by 30–50%. Hence, the modelled concentrations depend strongly on the coefficients a and g, and model users have to take care about a proper selection of them.
5. Conclusions and final remarks Due to the comprehensive evaluation of GRAL 3.5 using field experiments performed at five different road tunnels in Japan and Austria, it is felt that the model is able to give reasonable estimates for pollutant concentrations in the surroundings of tunnel portals. Nevertheless, there are currently two empirical parameters in the model, which had to be adjusted according to differences
regarding traffic volume and construction peculiarities of the different tunnel sites used for model evaluation. As it was not possible by now, to establish a fixed rule on how to choose those parameters for a certain tunnel portal, the situation is discontenting at the moment. Although the new introduced stiffness parameter (Table 1) should guide users of the model to choose suited values and obtain therefore reasonable concentration distributions for other tunnel sites as well. GRAL 3.5 has not been extensively tested when temperature differences between tunnel air and ambient air become negative. This might be often the case during summer days. Thus, care has to be taken regarding such situations, as one can expect high pollutant concentrations due to reduced vertical mixing.
Acknowledgements The project was funded by the Austrian ministry for traffic, innovation and technology (No. 3.248). Air quality observations were supervised by H. Puxbaum (Technical University of Vienna) and performed by R. Ellinger (Laboratory for Environmental Analytics, Vienna). Meteorological data were recorded by E. Mursch-Radlgruber (University of Natural Resources, Vienna).
References Oettl, D., Sturm, P.J., Bacher, M., Pretterhofer, G., Almbauer, R.A., 2002. A simple model for the dispersion of pollutants from a road tunnel portal. Atmospheric Environment 36, 2943–2953. Oettl, D., Sturm, P.J., Pretterhofer, G., Bacher, M., Rodler, J., Almbauer, R.A., 2003a. Lagrangian dispersion modeling of vehicular emissions from a highway in complex terrain. Journal of the Air and Waste Management Association 53, 1233–1240. Oettl, D., Sturm, P.J., Almbauer, R.A., Okamoto, S., Horiuchi, K., 2003b. Dispersion from road tunnel portals: Comparison of two different modelling approaches. Atmospheric Environment 37, 5165–5175. Okamoto, S., Hada, Y., Konno, Y., Kobayashi, K., Horiuchi, K., 2001. Evaluation of the JH air quality simulation model for tunnel portals. Proceedings of the Seventh International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, 28–31 May, 2001, Belgirate, Italy, pp. 229– 233. Puxbaum, H., Ellinger, R., Greßlehner, K.H., Mursch-Radlgruber, E., Oettl, D., Staudinger, M., Sturm, P.J., 2003. Measuring and modelling pollutant dispersion at tunnel portals. Austrian Ministry for Traffic, Innovation and Technology, Rep. No. 532, p. 78.
Short communication
Evaluation of GRAL for the pollutant dispersion from a city street tunnel portal at depressed level Dietmar Oettl*, Peter Sturm, Raimund Almbauer Institute for Internal Combustion Engines and Thermodynamics, Graz University of Technology, Inffeldgasse 21, 8010 Graz, Austria Received 31 May 2004; accepted 23 June 2004
Abstract A new data set for evaluation of atmospheric dispersion models, designed to treat road tunnel portals, is presented. Further, the performance of GRAL 3.5 (Graz Lagrangian model Version 3.5) using this data is discussed. One main drawback of the model in its current version is that two empirical parameters have to be adjusted for each tunnel site separately in dependence on average traffic volume and construction of the tunnel portal. As the model has been tested for five different tunnel sites by now, which differ significantly in traffic volume, terrain, and construction, model users should be able to choose reasonable values for those empirical parameters for other tunnel sites as well. GRAL has not been tested against field data, where the tunnel air is colder than ambient air. Here, further research is necessary. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: GRAL; Tunnel portal; Dispersion; Tunnel jet; Lagrangian model
Software availability GRAL: Free of charge from the authors for the operating systems Windows XP and Windows 2000 Professional.
1. Introduction Often difficulties occur regarding the optimum location of tunnel portals, as in many cases the whole vehicle emissions inside the tunnel are concentrated there. Dispersion models have to be applied to evaluate air quality standards at nearby residential areas. Dispersion of pollutants from tunnel portals requires new approaches in modelling as it differs significantly compared to dispersion of pollutants from other sources * Corresponding author. Tel.: C43 316 873 8081; fax: C43 316 873 8080. E-mail address: [email protected] (D. Oettl). 1364-8152/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.envsoft.2004.06.001
(e.g. line sources). The polluted air streaming out of the tunnel portal (hereafter called tunnel jet) is strongly influenced by the ambient wind in a way, that it is being forced to change the direction according to it. As was found by Oettl et al. (2002), the ambient wind direction fluctuations force the tunnel jet also to change its position, and in this way forming a very effective dispersion process. For convenience, this dispersion process will be abbreviated by the acronym ADAPT (Additional diffusion by the influence of a fluctuating ambient wind field on the position of the tunnel jet). In addition, buoyancy may also be an important factor for the dispersion of pollutants from tunnel portals, due to temperature differences between the tunnel air and the ambient air (Okamoto et al., 2001). In order to have a well suited dispersion model based on the current understanding of the particular processes described above, a novel method has been developed in the frame of a research project funded by the Austrian science fund (No. 14075-TEC). The new approach is
500
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
described in Oettl et al. (2002) and is incorporated as an individual module in the model GRAL (Graz Lagrangian Model, Oettl et al., 2003a). The module has been tested against experimental data from four different tunnel portals, namely the Enrei, Hitachi, Ninomiya tunnel in Japan (Oettl et al., 2003b), and the Ehrentalerbergtunnel in Austria (Oettl et al., 2002). The Japanese tunnels are all located in complex terrain, while the Austrian tunnel is surrounded by rather flat topography. Meteorological conditions observed during the experiments showed a wide variation of wind speeds (0.6– 6.2 m s1), atmospheric stabilities (stable–unstable), and wind directions, where it was able to evaluate the model for a wide range of angles between tunnel jets and ambient winds including head winds. In the year 2000 another research project was launched by the Austrian ministry for traffic, innovation and technology. The project aimed at investigating the pollutant dispersion in the proximity of a city tunnel in Vienna (Kaisermuehlentunnel), where the portal is situated 5 m below the surroundings. This is a particular difference in construction compared to all the other tunnel sites studied by now, which has an effect on the dispersion. Also, the traffic volume found in the Kaisermuehlentunnel is much higher than that for the other tunnels mentioned above. In the frame of this project, four different models (ADMS, LASAT, MUMO, and GRAL) were compared (Puxbaum et al., 2003). On the basis of this comparison, it was decided to recommend GRAL for dispersion modelling from road tunnel emissions in Austria in a new national guideline in elaboration.
2. Experimental set-up at the Kaisermuehlentunnel The Kaisermuehlentunnel in Vienna (Austria) has a length of 2150 m and two bores for each direction with six lanes in total. The pollution dispersion was studied at the south-east tunnel portal, where there exist two additional lanes to exit or enter the highway (Fig. 1). Continuous air quality observations were performed at five locations 2.5 m above ground level. Among the various chemical species recorded, NOx was found to be most related with the traffic on the highway and the tunnel jet. Another sampling point was set-up inside the tunnel to determine the emissions from the tunnel portal. The necessary volume flux was derived by recording the flow velocity in the tunnel using a cup anemometer. The method was validated with additional tracer tests by means of N2O releases inside the tunnel in five cases. Both methods agreed within G10%. The meteorological data used for modelling are based on wind observations with a cup anemometer on a 10 m mast at site M1 (Fig. 1). The experimental investigation lasted over a period of 10 months, where data were recorded on a half-hourly basis. As considerable background concentrations for NOx were expected, two distinct meteorological conditions were considered for the model evaluation: first, wind directions between 230 and 275 (979 cases) and second, wind directions between 95 and 125 (826 cases). In the first case the background concentration of NOx could be determined by use of sampling point M1, and in case of easterly winds, the average of sampling points M2–M5 was taken as background concentration.
Fig. 1. Orthophoto of south-east portal of the Kaisermuehlentunnel in Vienna (Austria). M1–M5 denote the sampling points for NOx. Meteorological data were recorded on a 10 m mast at site M1 using a cup anemometer.
501
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
Altogether 1805 cases were selected from the database for the simulations. The average wind speed found during the selected meteorological situations was 3.1 m s1. The maximum and the minimum wind speeds were 10.2 m s1 and 0.1 m s1, respectively. Temperature differences between the tunnel air and the ambient air were observed between 9.1 K and C14.2 K, and exit velocities of the tunnel jet ranged between 0.9 m s1 and 6.5 m s1. The mean NOxemission at the portal was 3.9 kg h1. In order to accurately model the NOx-concentrations, it is necessary to take into account the NOx-emissions resulting from the lanes out and into the tunnel. For the lanes out of the tunnel the corresponding NOx-emissions were determined by the NOx-emission at the portal divided by the length of the tunnel. As traffic data were not available on an half-hourly basis, NOx-emissions for the lanes into the tunnel were roughly estimated by assuming the same amount of traffic as out of the tunnel. This assumption clearly increases the uncertainty regarding the modelled mean half-hourly concentrations, while one can expect that it is a good estimation for the average concentrations over the whole period. The background concentrations for NOx were 35 mg m3 for easterly wind directions and 25 mg m3 for westerly winds.
3. Model set-up The dispersion model GRAL 3.5, respectively the tunnel module part of it, is described in detail in Oettl et al. (2002). It was already suggested in Oettl et al. (2002) to adjust some empirical parameters in the model to account for conditions, which differ in terms of construction and traffic volume from those, where the model has been validated by now. For the Kaisermuehlentunnel, the vertical dispersion has not been changed, while the ADAPT has been modified. The horizontal position of the tunnel jet is modelled by simulating the along and cross wind component of the jet, which depends on the ambient wind.
b, g the empirical constants; t the dispersion time [s]; UnS the cross wind component of the tunnel jet [m s1]; and UnA the ambient wind component perpendicular to the tunnel jet [m s1]. In case of the Ehrentalerbergtunnel, which has a rather low traffic volume of only a few hundred vehicles per hour, the parameters a and g were set equal to 1 (Oettl et al., 2002), while in case of the three Japanese tunnel sites, which have a traffic volume of about 1000–1400 vehicles per hour, a and g were set equal to 0.50 (Oettl et al., 2003b). The Kaisermuehlentunnel has a traffic volume of about 3740 vehicles per hour and furthermore, the tunnel portal is situated some 5 m below the surface. Thus, it is not surprising, that the ADAPT can be expected to be much lower compared to the other tunnel sites mentioned before. Therefore, a value of 0.003 (almost no change of the tunnel jet orientation) has been used for a and g in the simulations, except for head winds (south-easterly winds), where a value of 0.60 was chosen in order to obtain reasonable results. The motivation to take a higher value in case of head winds came from the consideration, that the ambient wind might not be influenced by the construction as it is the case for all other wind directions. It is apparent, that the values for a and g can vary over a wide range. Thus, it is quite difficult for users to choose those parameters properly. In order to make it easier for users to specify the values, a so-called stiffness parameter has been introduced, which is related with the parameters a and g as shown in Fig. 2. Table 1 provides guiding values for the stiffness parameter in dependence on traffic volume and construction of the tunnel portal. The stiffness parameter can take values between 0 and 100% depending on the traffic volume and construction of the tunnel portal. In order to get an estimate for the roughness length, sonic anemometer data were available to determine the
1 0.9
ð1Þ
KZað1CtÞ
ð2Þ
dUnS 1 2 Z bUnA 2 dt
ð3Þ
bZgð1CtÞ
ð4Þ
0.8
Parameters α and γ
dUp v2 Up ZK 2 dt vy
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
20
40
60
80
100
Stiffness parameter [%]
where Up is the along wind component of the tunnel jet [m s1]; K the turbulent exchange coefficient [m2 s1]; a,
Fig. 2. Relation between the stiffness parameter and the empirical parameters a and g.
502
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
Table 1 Guiding values for the stiffness parameter for a tunnel jet Tunnel
Stiffness parameter
Wind direction
Construction
Traffic [veh./h]
Terrain
Enrei
20–50 70
Head wind Other directions
At grade
w1300
Complex
Hitachi
20–50 70
Head wind Other directions
At grade
w1000
Complex
Ninomiya
20–50 70
Head wind Other directions
At grade
w1250
Complex
Ehrentalerberg
0 0–20
Head wind Other directions
5 m below the surface
w500
Flat
Kaisermuehlen
0–30 90
Head wind Other directions
5 m below the surface
w3800
Flat
friction velocity and subsequently the roughness length using u) z uðzÞZ ln ; ð5Þ z0 k where u(z) is the ambient wind speed depending on the height above ground in m s1, u* is the friction velocity as observed by the sonic anemometer in m s1, k is the Ka´rma´n constant taken to be 0.35, z is the height above ground level in m, and z0 is the roughness length in m. Eq. (5) is valid for neutral conditions only. A value of 0.17 G 0.08 m was obtained from the sonic anemometer recordings for the roughness length. Simulations were performed with a 5 m ! 5 m horizontal grid, and a vertical spacing of 1 m.
4. Results As most of the dispersion models, the tunnel module in GRAL 3.5 is based on classified average turbulence conditions. This means that relatively large uncertainty arises concerning modelled concentrations compared with observed ones paired in space and time, while average concentrations at fixed receptors will be more accurately modelled. Fig. 3 depicts the modelled average NOx-concentrations for the westerly winds and Fig. 4 shows the same but for easterly winds. The mean concentrations observed for the westerly winds were 38 mg m3, 23 mg m3, 60 mg m3, and 147 mg m3 for monitoring stations M2, M3, M4 and M5, respectively. It is
Fig. 3. Mean simulated concentrations for NOx for westerly winds in mg m3.
503
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504
Fig. 4. Mean simulated concentrations for NOx for easterly winds in mg m3.
elled. Fig. 5 depicts the cumulative frequency distribution of the observed and modelled NOx-concentrations at site M5 for the westerly winds. As the model is based on average turbulence conditions, one cannot expect that it predicts the highest or lowest percentiles of concentration distributions perfectly. It is therefore not surprising that an underestimation of higher percentiles and an overestimation of lower percentiles was found, while the median value was modelled reasonably. Fig. 6 shows the same as Fig. 5 but for the monitoring site M1 for easterly winds. At this site a clear underestimation of the median concentration is obtained with GRAL 3.5. This might be regarded as an improper treatment of head wind situations, disturbances of the tunnel jet due to the portal itself under such conditions, or to the sharp variation of concentrations found near M1.
1.0 0.9
Cumulative frequency
remarkable, that M2 shows higher concentrations than M3 although for westerly winds it was expected that M2 should be less effected by the tunnel and road emissions than M3. Most probably the reason for the somewhat higher concentration at M2 is due to traffic emissions on the smaller streets adjacent to M2. These are not included in the background concentrations observed at M1. There is also a smaller street passing by the stations M3–M5, but it has lower traffic than all the small streets together crossing close to M2. Modelled NOx-concentrations with GRAL 3.5 give 8 mg m3, 21 mg m3, 60 mg m3, and 128 mg m3 at sites M2–M5. A rather good agreement was obtained for all monitoring sites except for M2, where the model suggested a substantial lower concentration than observed due to the reason mentioned above. In case of easterly winds, the modelled concentration at M1 was 110 mg m3 compared to observed 120 mg m3. The largest deviation between the average modelled and observed concentration was therefore about 13%, except at M2. Simulations were also performed for the tunnel jet alone without including the lanes into and out of the tunnel. In this way the contribution of the tunnel jet to the observed concentrations could be estimated. At sites M1 and M5 the contribution of the tunnel jet to the recorded mean concentrations (without background concentration) was about 52% and 60%, while at the monitoring site M3 the contribution was only about 10%. This circumstance is one of the main drawbacks of the data set, as it is not possible to evaluate the model performance for the tunnel module alone, but the roads have also to be considered. For practical purposes, it is also of interest how high percentiles of concentration distributions can be mod-
0.8
Observed GRAL3.5
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0
200
400
600
800
1000
Concentration [µg/m3] Fig. 5. Cumulative frequency distribution for the observed and modelled NOx-concentrations at site M5 for westerly winds.
504
D. Oettl et al. / Environmental Modelling & Software 20 (2005) 499–504 1.0
Cumulative frequency
0.9 0.8 Observed GRAL3.5
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0
100
200
300
400
Concentration
500
600
700
[µg/m3]
Fig. 6. Cumulative frequency distribution for the observed and modelled NOx-concentrations at site M1 for easterly winds.
In our opinion, a statistical evaluation of the model performance when comparing half-hourly observed and modelled concentrations at each site is not very helpful, as has already been outlined, the model is not designed for such applications. Furthermore, in this study, there exists relatively large uncertainty regarding modelled concentrations on a half-hourly basis due to the lack of knowledge of the actual traffic emissions on the lanes into the tunnel. Finally, a sensitivity analysis was made in setting the empirical coefficients a and g equal to 0.012, a value which is somewhat between the ones chosen for head winds and all the other wind directions before. The resultant modelled concentrations were 87 mg m3, 135 mg m3, 106 mg m3, and 52 mg m3 at the monitoring sites M3–M5 and M1. While concentrations at M3 and M4 were substantially overestimated by more than a factor of two, concentrations at M5 and M1 were underestimated by 30–50%. Hence, the modelled concentrations depend strongly on the coefficients a and g, and model users have to take care about a proper selection of them.
5. Conclusions and final remarks Due to the comprehensive evaluation of GRAL 3.5 using field experiments performed at five different road tunnels in Japan and Austria, it is felt that the model is able to give reasonable estimates for pollutant concentrations in the surroundings of tunnel portals. Nevertheless, there are currently two empirical parameters in the model, which had to be adjusted according to differences
regarding traffic volume and construction peculiarities of the different tunnel sites used for model evaluation. As it was not possible by now, to establish a fixed rule on how to choose those parameters for a certain tunnel portal, the situation is discontenting at the moment. Although the new introduced stiffness parameter (Table 1) should guide users of the model to choose suited values and obtain therefore reasonable concentration distributions for other tunnel sites as well. GRAL 3.5 has not been extensively tested when temperature differences between tunnel air and ambient air become negative. This might be often the case during summer days. Thus, care has to be taken regarding such situations, as one can expect high pollutant concentrations due to reduced vertical mixing.
Acknowledgements The project was funded by the Austrian ministry for traffic, innovation and technology (No. 3.248). Air quality observations were supervised by H. Puxbaum (Technical University of Vienna) and performed by R. Ellinger (Laboratory for Environmental Analytics, Vienna). Meteorological data were recorded by E. Mursch-Radlgruber (University of Natural Resources, Vienna).
References Oettl, D., Sturm, P.J., Bacher, M., Pretterhofer, G., Almbauer, R.A., 2002. A simple model for the dispersion of pollutants from a road tunnel portal. Atmospheric Environment 36, 2943–2953. Oettl, D., Sturm, P.J., Pretterhofer, G., Bacher, M., Rodler, J., Almbauer, R.A., 2003a. Lagrangian dispersion modeling of vehicular emissions from a highway in complex terrain. Journal of the Air and Waste Management Association 53, 1233–1240. Oettl, D., Sturm, P.J., Almbauer, R.A., Okamoto, S., Horiuchi, K., 2003b. Dispersion from road tunnel portals: Comparison of two different modelling approaches. Atmospheric Environment 37, 5165–5175. Okamoto, S., Hada, Y., Konno, Y., Kobayashi, K., Horiuchi, K., 2001. Evaluation of the JH air quality simulation model for tunnel portals. Proceedings of the Seventh International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, 28–31 May, 2001, Belgirate, Italy, pp. 229– 233. Puxbaum, H., Ellinger, R., Greßlehner, K.H., Mursch-Radlgruber, E., Oettl, D., Staudinger, M., Sturm, P.J., 2003. Measuring and modelling pollutant dispersion at tunnel portals. Austrian Ministry for Traffic, Innovation and Technology, Rep. No. 532, p. 78.