The Representation of Urban Areas in Air Quality Models
Chapter 9
The Representation of Urban Areas in Air Quality Models: Validation in the Basel Region (Switzerland)
R. Hamdi[1]and G. Schayes
Royal Meteorological Institute, Brussels, Belgium
Institute of Astronomy and Geophysics, University of Louvain, Belgium
Abstract
Pollution increases in urban areas due to traffic, and local industry has become a major public health issue. The concentration of pollutants in urban areas are largely conditioned by the meteorological processes, which control the time and spatial scales of their horizontal and vertical dispersions. These processes are in turn modified by the presence of a city, especially in the Atmospheric Boundary Layer (ABL) (see for example the definition given in Stull [1984]). Even through it represents only a small part of the atmosphere (its depth ranges between few hundreds of meters and few km), this is the layer where most air pollutants are confined. However, the state of the art in most urban dispersion models is still to use turbulence and surface exchanges parameterizations, which are designed for non-urban terrain, partially with slight urban adjustments, but without taking into account the effects of the extremely rough surfaces of cities [Hanna et al. 1993; Chang and Hanna 2004]. Cities affect the local weather by perturbing the wind, temperature, moisture, turbulence, and surface energy budget field [Roth 2000]. In this study, we use the urban parameterization scheme of Martilli et al. [2002], which is a recent example of the drag force approach for momentum and turbulent kinetic energy. This scheme computes the impact of every urban surface type (Roof, Road, Wall) on the momentum, heat, and turbulent kinetic energy equation separately, these additional terms are taken into account in proportion to the area of their respective surface fractions. It also considers the shadowing and radiation trapping effect that consequently influence the calculation of the turbulent length scales. Results are compared with simulations using the classical way of parameterizing urban area effects on the one hand and with measurements within and above a street canyon on the other hand. Measurements were obtained during the intensive observation period of the Basel Urban Boundary Layer Experiment (BUBBLE), which is probably the most detailed European urban boundary layer experiment with a number of experimental activities in the city of Basel (Switzerland).
1. Introduction
Nowadays, pollution increases in urban areas due to traffic, and local industry has become a major public health issue. The concentration of pollutants in urban areas are largely conditioned by the meteorological processes, which control the time and spatial scales of their horizontal and vertical dispersion. These processes are in turn modified by the presence of a city, especially in the Atmospheric Boundary Layer (ABL). The need for an accurate modeling of the atmospheric flow field over urban areas has been pointed out in earlier studies. The aim of this chapter is to quantify the effects of a major urban agglomeration, in this case Basel, on the atmospheric boundary layer by means of a numerical model.
According to Wanner and Hertig [1984], local and regional wind systems play an important role for the climate of Swiss cities embedded in complex terrain. For this reason, the climate in the region of Basel is not only influenced by urban structures but also largely by terrain-induced effects.
The urban parameterization scheme of Martilli et al. [2002] is implemented in a new version of our mesoscale meteorological model (Topographic Vorticity-mode Mesoscale model) TVM called ''TVM perturbation mode''. TVM can be forced now by meteorological data, from a large-scale model using a perturbation mode [Genikovich and Schayes 2007], for better adaptation to the local weather situation.
In the present chapter, TVM is applied to the situation over the region of Basel using observations made in the context of (Basel Urban Boundary Layer Experiment) BUBBLE project. The objective is that of running the full three-dimensional mesoscale model including the urban scheme and the new perturbation mode over a realistic domain while considering the topography and land coverage of this domain. Results calculated by the mesoscale model for turbulent fluxes and meteorological variables are compared with results obtained with the classical approach using Monin-Obukhov Similarity Theory (MOST) formulation.
2. Numerical Models and Land-Use Distribution in the Basel Region
2.1. TVM Perturbation Mode
The Three-dimensional Topographic Vorticity-Mode Mesoscale (TVM) numerical model solves the atmospheric dynamic equations in vorticity mode and uses non hydrostatic, Boussinesq and anelastic approximations [Schayes et al. 1996; Thunis and Clappier 2000]. A constant flux surface boundary layer based on the Monin-Obukhov Similarity Theory makes the connection with the ground surface. Surface temperature and moisture values are computed using surface energy and moisture balance equations with a modified force restore model of Deardorff [1978] forced by the solar radiation value. The turbulence scheme is 1.5-order closure following the formulation of Therry and Lacarrère [1983].
In order to run TVM in its basic state, users must provide the model with a vertical profile of temperature, wind, and humidity as the initial conditions. However, for better adaptation to the local weather situation (regional wind systems) prevailing the entire simulation period, the model can be forced now by meteorological data from a large-scale model using a perturbation mode (for details see Genikovich and Schayes [2007]). The perturbation version of TVM is build upon the partitioning of each of the prognostic meteorological variables into a large-scale contribution (L) and a meso-local scale contribution (M):
(1)
and make use of the Boussinesq approximation. The large-scale flow is assumed to be in equilibrium, non turbulent and in hydrostatic equilibrium. The new governing equations for the TVM perturbation mode are described in Genikovich and Schayes [2007] and in the present formulation, the boundary conditions over any mesoscale perturbation variable, whererepresent the components of the wind speed andthe temperature, is specified as follows:
- Upper boundary conditions: perturbation.
- Lateral boundary conditions:,
- Bottom boundary conditions: for wind,and for temperature, an ordinary ground energy balance is performed onand then.
The perturbation variable is also forced to decrease exponentially with a time constant of about 4 to 12 hours, decreasing with height above ground.
The large-scale informations used here is provided by the ERA-40 re-analysis data sets from the European Centre for Medium-rang Weather Forecasts (ECMWF). These data are provided at 6-hour intervals, with a horizontal resolution of 1° (approximately 100 km). It specifies the profiles of windand temperaturenear the centre of the domain under study. For the present level of approximation, we need the profiles of temperature and wind speed of the centre of the meso domain only and these profiles are supposed to be valid for the whole meso domain. This approximation is supposed to be valid if the domain size is not too large, as is in the present chapter.
2.2. Parameterization of Urban Land Use
The principle of the urban surface exchange parameterization of Martilli et al. [2002], is that extra terms are added to the momentum, heat, and turbulent kinetic energy equations separately. These terms are taken into account in proportions to the area of the surface fraction of three urban surfaces type: road, wall, and roof. So, no moisture flux is allowed in the urban canyon. However, in the literature a lot of numerical simulations and field measurements indicate that increasing vegetation cover can be effective in reducing the surface and air temperature near the ground [Taha 1999]. Eliasson [1996] found that the air temperature difference observed between the large park and the city center is of the same order as the average urban-rural air temperature difference. In order to take into account the vegetation effect on urban canopy, we divide the urban grid cell into an non-urban fraction (vegetated fractionV) and an urban fraction 1-V, and then further subdivide the urban canopy fraction into road, wall, and roof according to Martilli's scheme [Hamdi and Schayes 2008].
2.3. Numerical Domain and Land Cover
The numerical domain includes 55x55 km covering the city of Basel and itsimmediate surroundings, as well as a small part of the Black Forest and Jura Mountains. The horizontal resolution is set to 1 km. Due to this finer resolution altitude reach 1200 m a.s.l in the Black Forest (see Figure 1). Topographic data were taken from a Digital Elevation Model (DEM) with a horizontal resolution of 120 m, covering the whole Upper Rhine valley from Strasbourg (France) to Basel. The data were collected during the RECLIP project [Scherer et al. 1994]. The TVM model extended vertically up to 5000 m above ground level with a vertical resolution from 20 m close to the ground to 500 m at the top of the domain with a total of 33 layers.
Figure 1. Topography of the domain.
Table 1 Description of the 14 classes defined in REKLIP data base
Code / land-use (name)1 / Water
2 / High density settlement
3 / Medium density settlement
4 / Low density settlement
5 / Industrial areas
6 / Coniferous forest
7 / Mixed forest
8 / Deciduous forest
9 / Fallow land
10 / Gardens
11 / Viniculture
12 / Grassland
13 / Winter fields
14 / Spring fields
The land use classification of the Basel region is based on the Landsat-Data during REKLIP project, which covers about the whole upper Rhine Valley with an 30 m resolution and 14 classes (see Table 1). Using this finer resolution of the land cover data set (30 m x 30 m), we have calculated for each TVM ground-level cells (1km x 1km), the maximum percentage of all land use classes. The land cover for TVM cell will be equal to the land cover corresponding to the maximum one (see Figure 2).
3. Observational Sites
Between 10 June and 10 July 2002, an intensive observation period (IOP) was carried out in the region of Basel (BUBBLE), which embedded many activities from international research groups. The overall framework and the experimental activities during BUBBLE are documented in Rotach et al. [2005]. The BUBBLE data set involves 30 experimental or permanent sites from the greater Basel area (See Table 4.4 in Roulet [2004]). The map in Figure 3 shows the topography and setting of the experimental activities in the city of Basel.
Figure 2 Land use categories considered by TVM. Grid cells designated as urban areas are 2, 3, 4, and 5. Refer to Table 1 for definitions of these subcategories.
Figure 3. Map of all sites operated during the BUBBLE IOP in June/July 2002.
Not all stations were used for the validation. In the present chapter, we will compare model results with data from stations representing three types of land use: urban, suburban, and rural. For this purpose, 3 stations were chosen (more details can be found in Christen [2005]):
- Basel-Sperrstrasse U(Ue1 on the map): This site is located in a heavily built-up part of the city (''European urban'', dense urban, mainly residential 3 to 4 storey buildings in blocks, flat commercial and light industrial buildings in the backyards). The area is characterized by a low vegetation cover and therefore low moisture availability (see Table 2). The measurements set up consists of a tower inside a street canyon reaching up to a little more than two times the building height, where six ultrasonic anemometer-thermometers and full radiation component measurements are installed.
- Allschwil-Rämelstrasse S(Se1 on the map): During summer 2002 (IOP) a pneumatic tower was installed with 3 levels of sonic anemometers mounted above a flat gravel roof and full radiation components at tower top over this ''European suburban''. The site consists of a relatively homogeneous city part with older single and row houses and small streets. In the backyard there are many trees and scrubs.
- Grenzach R (Re1 on the map): The rural surface energy balance station R is located 5 km East of the city inside the plain of the river Rhine. The land use is mainly agriculture (non-irrigated grassland and crops). A sonic was mounted on an existing 110 m radio tower at a height of 28 m. A fast hygrometer at same height provided information on humidity fluctuations. Additionally, surface measurements were carried out 100 m north of the tower base over grassland, where all radiation components and soil measurements were monitored at 1.4 m. The turbulence and surface energy balance measurements were operated from April 24, 2002 until July 12, 2002.
Table 2. Description of the micro-meteorological towers at the urban and suburban sites
Site / Basel-Sperrstrasse (U) / Allschwil-Rämelstrasse (S)Location / 47° 33' N
7° 35' E / 47° 33' N
7° 33' E
Height a.s.l / 255 m / 277 m
Height of tower / 32 m / 16 m
Height/Width (H/W)a / 1.30 / 0.63
Vegetationa / V=16% / V=53%
Table 2. (Continued)
Site / Basel-Sperrstrasse (U) / Allschwil-Rämelstrasse (S)Ultrasonic Anemometers / F31.7 m
E22.4 m
D17.9 m
C14.7 m
B11.3 m
A3.6 m / C15.8 m
B12.1 m
A8.3 m
Net radiation / 31.5 m / 15.1 m
Latent and Sensible heat / 31.7 m / 15.8 m
aLocal value, calculated for a circle of 250 m around the site.
4. Set up of the Simulations
Two simulations were performed: the first simulation, denoted ''urban'', uses the urban version of TVM, the second simulation, called ''class'', represents the classical approach used in TVM to account for urban surface using Monin-Obukhov similarity theory (MOST) (city characterized only by a change in roughness length and the surface conditions).
4.1.Classical Simulation
In the simulation that uses the classical roughness approach, the urban areas are assumed to be homogeneous (based on land use) and are characterized with the same physical attributes (see Table 3), which are also defined based on dominant land use category.
4.2. Urban Simulation
The percentage of vegetation Vto be specified in each cell and needed for the weighting of the urban and rural contributions was determined from the REKLIP data base. In fact, we have calculated for each TVM ground-level cells (1km 1km), the percentage of each class and then the maximum of all percentages in order to find the dominant land use category. If the dominant land use category corresponds to one of the four urban subcategories (2, 3, 4, 5) then the fraction of vegetation inside the urban ground-level cell will be the sum of the percentages of all classes from 6 to 14 which correspond to various vegetation land use. In Figure 4, we have calculated the frequency distribution of the fraction of vegetationVfor two urban category: ''High density settlement'' and ''Low density settlement''.
Table 3 Land use parameters used for the classical simulation
Soil/vegetation type / Albedo / Emissivity / Stomatal resistance(s m-1) / Heat capacity(J m-2 K-1) / Roughness length(m)Water / 0.03 / 0.99 / 0 / 1.00E+00 / 0.001
High density settlement / 0.15 / 0.95 / 300 / 5.00E+04 / 1.000
Medium density settlement / 0.16 / 0.95 / 300 / 5.00E+04 / 1.000
Low density settlement / 0.17 / 0.95 / 300 / 5.00E+04 / 1.000
Industrial areas / 0.17 / 0.95 / 300 / 5.00E+04 / 1.000
Coniferous forest / 0.12 / 0.98 / 100 / 2.00E+05 / 1.500
Mixed forest / 0.14 / 0.98 / 100 / 2.00E+05 / 1.500
Deciduous forest / 0.17 / 0.98 / 100 / 2.00E+05 / 1.500
Fallow land / 0.17 / 0.98 / 100 / 3.40E+05 / 0.100
Gardens / 0.15 / 0.98 / 100 / 2.00E+05 / 0.100
Viniculture / 0.17 / 0.98 / 50 / 2.00E+05 / 0.100
Grassland / 0.20 / 0.98 / 50 / 3.40E+05 / 0.030
Winter fields / 0.20 / 0.98 / 50 / 2.00E+05 / 0.100
Spring fields / 0.17 / 0.98 / 50 / 2.00E+05 / 0.100
Figure 4. The frequency distribution of the fraction of vegetation V (%) for two urban subcategories: ''High density settlement'' and ''Low density settlement''.
Based on these results, three urban classes were defined in terms of specific building parameters (width, height, density) and the fraction of vegetation V(see Table 4):
- High density settlement (Urban): mean building height 15 m andV=15%.
- Medium density settlement (Urban): mean buildings height 8 m and V=30%.
- Suburban and industrial areas (Suburban): mean building height 6.5 m andV=55%.
The three urban classes defined above only differ in the distribution of building-height density and in street widths. Two perpendicular street orientations with angles of 70° and 160° were defined (identical in the three classes). The thermal characteristics of the building materials (see Table 5) were established using data from the literature, and refined after the validation of the urban module [Hamdi and Schayes 2008].
Table 4 Building-height distribution for the three urban classes defined
Buildings height (m) / Percentage of building (%)Class 1 / Class 2 / Class 3
6 / - / 25 / 75
8 / - / 50 / 25
10 / 25 / 25 / -
15 / 50 / - / -
20 / 25 / - / -
Roof size(m) / 15 / 15 / 15
Street size(m) / 15 / 20 / 30
Vegetation (%) / 15 / 30 / 55
Table 5Physical properties of urban elements. KSthe thermal conductivity of the material. CS the specific heat of the material. Tinterne the initial temperature of the material. the emissivity of the surface. the albedo. z0 the roughness length
KS (m2 s-1) / CS (J m-3 K-1) / Tinterne (K) / / / z0 (m)Wall / 7.10E-07 / 1.40E+06 / 293 / 0.90 / 0.14 / -----
Roof / 7.90E-07 / 1.40E+06 / 293 / 0.90 / 0.14 / 0.01
Road / 1.05E-06 / 1.80E+06 / 290 / 0.95 / 0.08 / 0.01
TVM does not consider any condensation or cloud-formation processes. Therefore, no feedback to the solar-radiation calculation (attenuation through cloud episodes) is taken into account in the model. Since this phenomenon plays an important role for the amount of energy available for turbulence generation near the ground, and should therefore be included, solar radiation can be prescribed to the model by data input (from the urban site U), rather than calculating it by the model itself. This constitutes an ''indirect'' consideration of cloud formation in the model and will improve the fit to local meteorological conditions prevailing during the episode of simulation. The period of the simulation extends from 25 June to 27 June 2002 and corresponds to a thermally driven meteorological situation included in the the second half of the IOP in the frame of the BUBBLE campaign. This period is a unique opportunity, firstly, to quantify the impact of the urban exchange parameterization on meteorological modeling and secondly to investigate its results over completely different urban surfaces under the same synoptic forcing.
5. Results
5.1. Energy Budget and Surface Meteorological Fields
5.1.1. Net Radiation
Figure 5. Ensemble diurnal course of net radiation Q* (W m-2) at the urban (U), suburban (S) and rural (R) sites from June 25 to June 27 according to the ''urban'' and ''class'' simulations, and according to observations from the BUBBLE experiment.
The comparison between the three sites shows that the maximum daily observed net radiation Q* is slightly greater in the urban site than in the suburban or the rural areas. The city center gains more energy compared with the suburban or the rural site, an effect mainly controlled by the lower effective albedo of the urban site[Christen and Vogt 2004]. The midday average differences are:and. Throughout the night, the urban sites lose more energy through Q*than the suburban or the rural sites:and. The three-dimensional geometry of the urban canyon induces opposite effects, which are considered by the urban module of Martilli et al. [2002]. On the one hand, trapping of the incident solar radiation and longwave emission of the surface. On the other hand, shade effects directly proportional to the aspect ratio H/W of the streets.