Summary

Automotive traffic in cities is a major contributor to the pollution level and has an important adverse effect on the air quality and therefore on health. Tools supporting environmental management are available to calculate air quality near individual sources but are lacking at the urban scale. The influence of tunnels, buildings and other obstructions on the wind flow pattern as well as the traffic induced pollution have to be taken into consideration. In this context, urban scales air quality management tools are developed to get a better understanding of physical and chemical processes in order to find effective solutions to air quality problems.

The present work focuses on modelling traffic induced flow field and turbulence, traffic emissions and turbulence dispersion in a canyon street and in a tunnel. In order to check the ability of the model to reproduce the flow induced by the wind around buildings, a 2D approach with three different geometrical configurations of buildings (i.e. flat or inclined roof) is implemented. A single pollutant source in the street-canyon is introduced in the model. The velocity and concentration fields simulated are compared with experimental data. Then, a 3D configuration with a tunnel is considered. Vehicle are considered as sources of additional momentum which is parameterized through the average vehicle speed and through geometrical parameters describing the traffic (i.e. average vehicle height and average distance between two consecutive vehicles). Source terms of turbulence are also added. Pollution induced by the traffic is taken into account by considering surface or volume sources of heat, momentum and mass on the road. The strength of these sources is determined by a global mass, momentum and heat balance.

The numerical study is conducted using a finite CFD code based on an artificial compressibility approach. We use an isotropic k-e model which is assumed to be valid for neutrally stratified conditions. The model reckons the velocity field as well as the concentrations of pollutants. Numerical results of concentrations are compared to experimental data. Despite some discrepancies, general trends are correctly reproduced.

Aim of the research

The main goal is to simulate traffic induced pollutant concentrations in urban configurations. This requires the addition of source terms corresponding to momentum and turbulent kinetic energy brought by cars. Moreover, pollutants may undergo transformations which are to be included into the model.

Activities during the year

During the year 1998, we focussed on traffic induced turbulence and emissions modelling. Two cases for which data are available have been implemented and tested :

1. A 2D flow over a canyon street with different building roof shapes (flat or triangular)
2. A 3D flow in a tunnel taking account of additional source terms from the traffic.

This first approach only gives a coarse representation of traffic related inputs. A more refined model of traffic induced turbulence is then developed for case 2. The momentum source term controls the traffic induced air motion and is modelled from the vehicle speed, the mean vehicle shape coefficient and the traffic density. The source term of turbulent kinetic energy is set in such a way that resulting concentration profiles are approximately constant within each transversal tunnel section, in agreement with the experiments. Assuming that the flow is fully turbulent in the domain, the production of mechanical turbulence is supposed to be compensated by its dissipation, leading thus to a local equilibrium. This assumption allows the use of standard wall functions for the turbulent kinetic energy and its dissipation. NOx and CO concentrations are simulated by adding a mass source term in the equation of transport for these entities. This mass source term is a function of the emission factor, the traffic density and the tunnel length.

Principal results

Study case 1

From case 1, we assess the ability of the code to reproduce the velocity and concentration fields for specific geometries (see Fig. 1). Velocity fields reveal a clockwise vortex in the canyon for configurations 1 and 3 and an anticlockwise vortex for configuration 2. Comparison of the concentration profiles at the leeward and at the windward sides of the canyon street with the experimental data (Kastner-Klein et al., 1996) is made in Fig 2. For configuration 2, the maximum concentration is measured at the windward side in agreement with our prediction, confirming thus the direction of the vortex. For all configurations, the lower concentrations are reasonably predicted, however high discrepancies are noticed for the higher concentrations. We note in particular that the predicted concentrations on the leeward side in configuration 3 is quite different from that of configuration 1 although the experiments give similar concentrations level. General trends, however, are predicted by the model.

Study case 2

This test case present the pollution concentration predicted with our code for a traffic situation occurring on the 11/27/1991 in the Leopold II tunnel in Brussels. The tunnel is bidirectional of about 600m length, 14m width and 4.5m high. The inlet and outlet ramps are about 100m long. The station where pollutant concentrations are measured is situated at 300m from the tunnel inlet (Fig. 3). The neighbouring buildings are not represented and the traffic is modelled only in one direction for the tunnel. The mesh includes about 100000 points.

The inlet velocity profile is modelled by a logarithmic law with a roughness height of 0.1m. The temperature profile is taken as linear. Values of turbulent kinetic energy and turbulent dissipation are based on a turbulence level at 30m above the ground of 30% and a turbulent length scale of 50m. Background concentrations of pollutants are imposed at the inlet. CO Pollution induced by the traffic is specified through an emission factor CO(g/km/car) from which the corresponding mass source term for CO in the tunnel is :

CO(g/s)=CO(g/km/car)xL_tunnel(km)xQcar(cars/h)

where L_tunnel is the tunnel length and Qcar is the traffic density. The predicted NOx concentrations are obtained by a rescaling of the results obtained for CO :

NOx-NOx_background=MM_NOx/MM_COxQ_NOx/Q_COx(CO-CO_background)

where MM_Nox and MM_CO are the molecular masses in g/mole and Q_NOx and Q_CO are emissions factors in g/km/car. The distribution of CO concentration along the tunnel grows linearly from the inlet to the outlet and is almost constant in the measurement cross section. Fig. 4 shows the calculated and measured concentrations (IHE/LVM, 1990) of CO and NOx as a function of the traffic density. Although the measured data appear to be scattered, the predicted concentrations are in good agreement with the experimental data. The results look therefore reasonable although assumptions on the mean driving speed and on the emission data had to be made.

Main conclusions

The code has been applied for neutral atmosphere using the standard isotropic k-e model with wall functions. Special attention has been paid to the impact of specific geometries on the flow and pollutant distributions. Furthermore, the appropriate source terms have been added into the code to take account of the traffic interactions with the wind induced flow field. The equations of transport for CO and NOx have been solved simultaneously. Results lead to reasonable agreement with experimental data available.

Aim for the coming year

The main developments now should concern both the improvement of the hydrodynamics model and the addition of relevant chemical processes into the code.

The model in its present version is valid for a neutral atmosphere. For stable and unstable conditions, it is expected that an isotropic turbulence will fail in predicting accurate flow patterns. As the turbulence level strongly influence the dispersion of pollutants, the improvement of turbulence models is crucial. In the next stage, a non-linear algebraic Reynolds stress turbulence model taking account of turbulence anisotropy will be applied.

Other developments concern the implementation of chemical processes occurring in urban areas and the assessment of their impact. Chemical reactions (e.g. nitrogen, ozone and VOCs chemistry) and their interactions at the local scale will be investigated and parameterised. A module simulating particles transport and based on a Lagrangian formulation will also be tested for atmospheric pollution applications.

For the validation of the new modules, interactions with part (c) of the Saturn project (i.e. EXP) are strongly recommended. The ongoing work concerns the improvement and the validation of the formulation used for traffic induced turbulence. For the particular configuration of a street canyon, new sets of data are already available and will be compared to numerical results. The experimental facility consists of a wind tunnel where vehicles are simulated by small metal plates mounted on two belts moving along the modelled street canyon.

Acknowledgments

We particularly thank Dr P. Kastner-Klein for her contribution to the experimental part of the study.
 
 

References

IHE/LVM; Luchtverontreiniging in verkeerstunnels, Een evalutie van de verkeersemissies aan de hand van koolstofbalans, Meetcampagne november-december 1989, Technical report, Brussels (1990).

Kastner-Klein, P., E Fedorovitch and E Plate; Wind-tunnel case study of the atmospheric dispersion in the urban environment, Proceeding of the 4^th congress on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, Oostende 6-9 May, Belgium (1996).
 
 

Fig. 4 Comparison between predicted and measured CO and NOx concentrations as a function of the traffic density