Cityflux Perfluorocarbon Tracer Experiments

In June 2006, two perfluorocarbon tracer experiments were conducted in central Manchester UK as part of the CityFlux campaign. The main aim was to investigate vertical dispersion in an urban area during convective conditions , but dispersion mechanisms within the street network were also studied. Paired receptors were used in most cases where one receptor was located at ground level and one at roof level. One receptor was located on the roof of Portland Tower which is an 80 m high building in central Manchester. Source receptor distances in the two experiments varied between 120 and 600 m. The results reveal that maximum concentration was sometimes found at roof level rather than at ground level implying the effectiveness of convective forces on dispersion. The degree of vertical dispersion was found to be dependent on source receptor distance as well as on building height in proximity to the release site. Evidence of flow channelling in a street canyon was also found. Both a Gaussian profile and a street network model were applied and the results show that the urban topography may lead to highly effective flow channelling which therefore may be a very important dispersion mechanism should the right meteorological conditions prevail. The experimental results from this campaign have also been compared with a simple urban dispersion model that was developed during the DAPPLE framework and show good agreement with this. The results presented here are some of the first published regarding vertical dispersion. More tracer experiments are needed in order to further characterise vertical concentration profiles and their dependence on, for instance, atmospheric stability. The impact of urban topography on pollutant dispersion is important to focus on in future tracer experiments in order to improve performance of models as well as for our understanding of the relationship between air quality and public health.


Introduction
As part of the CityFlux campaign (Langford et al., 2009;C. L. Martin et al., 2009) two perfluorocarbon tracer experiments were carried out. These experiments, in conjunction with the REPARTEE perflurocarbon tracer experiments are part of the recent extensive work conducted regarding vertical dispersion in urban areas using perfluoro-5 carbon tracers (D. . In contrast to the REPARTEE tracer experiments that were conducted during neutral atmospheric stability, the CityFlux experiments took place during convective days.
Vertical dispersion has been a neglected area of research. A few tracer experiments were carried out in the mid twentieth century in mostly rural areas using more exotic 10 tracers such as smoke, aniline vapour and zinc cadmium sulphide (Sutton, 1947;Barad and Fuquay, 1962;Thompson, 1965Thompson, , 1966. In recent years a few articles have been published presenting results on vertical concentration profiles. Uniform vertical profiles in street canyons have been observed for source receptor distances of 700 m (Rotach et al., 2004) and 1 km (Cooke et al., 15 2000). In these experiments, receptors were placed on the roof of buildings with a maximum height of 30 m.
The most extensive campaign undertaken so far regarding vertical dispersion took place in Oklahoma City during the Joint URBAN 2003 campaign (Flaherty et al., 2007). 7 receptors were placed on a crane between 10 and 75 m above the ground at a dis-20 tance of 1 km away from the release site. Daytime experiments revealed that the plume was relatively well mixed where the lowest concentration returned was typically within 50% of the maximum concentration in the vertical profile. However, maximum concentration was returned more often in the lower half of the profile. A slightly less uniform profile was seen when night time experiments were undertaken. One experiment was 25 also carried out where the source receptor distance was only 500 m. Here, a distinct vertical concentration profile was found where the maximum concentration returned was closer to the ground than for the 1 km experiments. This article also presents results concerning the effect of channelling of flow in street canyons. A great deal of effort has been invested recently into understanding wind flows in street canyons and in particular its dependence on the above-roof wind flow (e.g. Arnold et al., 2004;Dobre et al. 2005;Wood et al., 2009). A street network model has been tested with the dataset here and reveals promise in order to understand 5 the close range dispersion in urban areas. In this model a plume encountering an intersection is divided up into two parts depending on the street alignment to the aboveroof wind direction.
The results presented in this article will also be compared with "the simple correlation for pollutant dispersion in the local environment" model that was developed as part of the DAPPLE framework (Neophytou and Britter, 2004). In this model the maximum concentration decays away with distance squared according to where C max is the maximum concentration downwind of a point source with a release rate (Q), U is the wind speed, x the source receptor distance, K a "constant" which is 15 dependent on urban morphology parameters such as the building frontal area (λ f ) and the building plan area (λ p ). The structure of the simple correlation model can be seen as a simplification of the Gaussian Plume formulation but here the concentration decay varies with the inverse square of the source receptor distance. In Eq.
(2) each term has been non-dimensionalised using H b which is the average 20 building height: Based on the results from the DAPPLE work a K of 10 or 20 was proposed to model maximum concentration following a release. This was proposed based on both experimental data from the DAPPLE area in central London as well as an investigation Birmingham by other research groups. The inverse square dependency on distance was also supported with wind tunnel experiments undertaken as part of the DAPPLE framework. The source receptor distances for all these were no longer than 10 km. Based on the results from the Joint URBAN 2003 experiments conducted in Oklahoma City the results showed that K was closer to 3 for daytime experiments and K close 5 to 10 for nighttime experiments (Hanna et al., 2007). The need to evaluate this model with an experimental dataset retrieved from other cities exists therefore in order to see what range K is in and what parameters K is dependent upon.

Experimental
The perfluorocarbon tracer release was made from pressurized cans containing Interactive Discussion of 293 • from north and an average wind speed of 5.7 m s −1 . The experiment was conducted during sunny conditions and the temperature was 18 • C.
The atmospheric background subtracted concentrations returned, normalised by the perfluorocarbon release rate, are shown in Table 1 and in Fig. 1. The average wind direction (green arrow in Fig. 1) reveals that there is a direct connection between the 5 source site (X) and Portland Tower and high perfluorocarbon tracer concentrations are found at the tower. Also tracer is found at receptor 4 and 5 but not at receptors 7-10 which can be used to estimate the size of the plume.

Vertical concentration profile
The vertical concentration profile at Portland Tower reveals that the receptor at ground 10 level (receptor 1) did not return the highest concentration but rather the receptor on the car park roof (receptor 3) slightly further away from the release site. The enhanced vertical dispersion can also be seen when comparing receptors 1 and 2. Receptor 2 at the top of Portland Tower returned a concentration slightly higher than that predicted by the Gaussian plume equation during stability class C, using Briggs' Interpolation 15 Formulae, shown in Table 1. This is the class preferred since the urban environment tends to favour neutral stability (Britter and Hanna, 2003). The unstable atmospheric conditions encountered here have caused the plume to extend in the vertical direction and the majority of the plume is located above ground level about 350 m away from the source.

Street channelling effect
The results from Receptor 1 and 4 (both at ground level) is used to evaluate the channelling effect. Both receptors returned similar amounts of tracer and the source receptor distances are similar but the source receptor angles are not. Receptor 1 is almost on the plume centreline based on the recorded wind direction at Portland Tower An attempt to quantify the channelling effect has been made by comparing this particular urban area with a flat terrain in a rural area where a Gaussian profile is assumed to approximate the lateral distribution of a plume. The concentration returned at recep-5 tor 4 would be 14% of that returned at receptor 1 when applying a Gaussian distribution with a plume centreline along the average wind direction recorded. Since the concentration returned for the two receptors are almost identical the urban topography has increased the tracer amounts at receptor 4 by at least a factor of 7. The profound difference of dispersion in urban areas compared with rural areas is therefore demon-10 strated here.
Another analysis in order to investigate the channelling effect is to compare the experimental results with the above-roof wind direction recorded relative to street geometry. In this analysis, the plume is assumed to split up into two components (R and 1-R) every time it encounters an intersection (Fig. 1). If all components are added to-15 gether this would imply a "concentration" at receptor 4 of (1-R) 5 and at receptor 1 of 5R (1-R) 4 . By equating these two equations R was determined to be 0.17. By dividing up the wind direction obtained at Portland Tower into vectors aligned as the street network R was calculated to 0.19, a remarkably close agreement. Street geometry and above roof wind direction can therefore explain the concentration profile found here within 20 the error of the measurements. The conclusion is therefore that the effect of street channelling has played a major role in plume dispersion. Interactive Discussion three samples (8 min each) were taken at each receptor sequentially, including a one minute gap in between every sample. No wind measurements from Portland Tower were available this day but the UK Met Office's weather station at Woodford airport reported an hourly (13:00-14:00) average wind direction of 200 • and an average wind speed of 3 m s −1 . The experiment was conducted during sunny conditions.

5
The atmospheric background subtracted concentrations returned, normalised by the perfluorocarbon release rate, are seen in Table 2 and Fig. 2 displays the results from the third sample. An elevation above atmospheric background concentrations are seen at all receptors. During the first and second sample only small perfluorocarbon tracer amounts were found at the sampling sites. However, during the third samples clear 10 vertical concentration profiles were found at several of the ground/roof paired receptor sites.

Vertical concentration profiles
During the first two samples only small amounts of the tracer reached any of the receptors. The concentrations returned for the receptors closest to the release site (11 and 15 12) are highly variable; in the first sample the ground level receptor returned the highest concentration and in the second the roof level receptor. This shows the intermittency of the dispersion of individual tracer parcels that arise during low wind speed conditions where it may be difficult to envisage a plume due to turbulence being the main mode of transportation rather than average wind speed. Therefore the tracer could be seen 20 to consist of "blobs" of air indicating their highly variable temporal and spatial scale (Martin et al., 2008).
During the time period when the third sample was taken, advection of the plume occurred towards most of the receptors. The paired receptors (7/8 and 9/10) both show a vertical gradient in each respective street canyon. For the paired receptors 7 25 and 8 the ratio ground to roof is 2:1 and for receptors 9 and 10 the corresponding ratio is 3:1. The difference in the results may be attributed to difference in receptor height. Although the source receptor distances for receptors 7/8 and 9/10 are similar to the source receptor distance for receptors 1 and 3 in the first experiment, the maximum concentration is found at ground level in the second experiment and not at roof level as in the first experiment. The difference in results can be explained by the height of the buildings in close proximity to the two release sites. The buildings located on the 5 northern side of Princess Street close to the release site in experiment 1 are generally 2 storey (except one 5 storey building). This would make it fairly easy for an above roof component of the plume to be formed thus resulting in higher concentrations returned at roof level rather than at ground level. It is more difficult for an above roof level component of the plume to be formed early on during the second experiment. Here, 10 the buildings on the northern side of Oxford Street close to the release site range between 4 and 6 storeys, that effectively trap the plume within the street canyon.

ACPD
At Portland Tower the vertical profile is inversed; the roof top receptor returns a concentration 42% higher than the ground level receptor implying the importance of enhanced vertical dispersion during convective situations. These results, in conjunction 15 with the profile obtained at Portland Tower during the first experiment, shows that a Gaussian profile is not a suitable description of dispersion during convective conditions since the maximum concentration was not found at ground level. The non-suitability of a Gaussian profile in the vertical was also demonstrated in the recent REPARTEE campaigns where it was found that the shape of the vertical profile (from ground level 20 up to 190 m) changed with source receptor distance (measured over 460-1370 m) 1 .

Street network model
The street network model has been tested against the results obtained from the second experiment for receptors 7 and 9 and for the third sample taken. The resulting component at receptor 7 was determined to 4R(1-R) 3 and at receptor 9 (1-R) 3 . By 25 taking into account the almost 20% higher concentration returned at receptor 7, R was determined to be 0.3. R calculated based from the wind direction obtained from Woodford airport and the street alignment is 0.52. The difference between the two values 35 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion of R is larger in experiment 2 than in experiment 1. However, the wind data is taken from Woodford and therefore the agreement is pretty reasonable, in keeping with wind analysis performed by Scaperdas and Colvile (1999). The model comparison made for the first experiment is very encouraging and suggests that "local" meteorological measurements are important in such an analysis.
5 5 Simple correlation model Figure 3 displays the results from both experiments compared with the simple correlation model. The wind speed from Portland Tower for the first experiment and from Woodford airport for the second experiment has been chosen. Figure 3 reveals that all experimental data points bar one are below the dotted line representing K =10. One 10 data point is located in between 10 and 20 in the very near field range which shows the large variability that may arise at close range. In this particular case the receptor returning high perfluorocarbon tracer concentration was located in the same street canyon as the release site which may give rise to very high tracer concentrations due to trapping in the street canyon. The same formulation and estimation of K , in a similar 15 manner to the results from the DAPPLE campaign in central London, UK, can therefore successfully explain the maximum concentration found during these experiments undertaken in central Manchester. The simple correlation model has here been evaluated over distances up to 600 m. The range of validity of this model has been under scrutiny and a suggested limit of 20 50 times the average building height was proposed during the DAPPLE framework. At longer distances the decay rate is expected to be reduced from an inverse square to inverse 1.5 or inverse 1.75. No sign of this is seen here but the source receptor distances are too short in order to investigate that (maximum source receptor distance is 600 m for these experiments). Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion

Conclusions
Vertical concentration profiles during convective conditions have been presented. Convective forces play an important part in pollutant dispersion where maximum concentration was sometimes encountered on the roof of a building or tower rather than on the ground. The possibility of formation of an above-roof component of the tracer re-5 leased is important in evaluation of vertical concentration profiles and the above-roof component is dependent on the building height close to the release site. The importance of street channelling was also shown and again reflects the importance urban topography exercise on dispersion. A simple street network model could explain the dispersion pattern found in the first experiment where a relatively uniform 10 street network is seen. The simple correlation model developed during the DAPPLE work has been accurately used to model the maximum concentration in these experiments. Introduction