Urban dispersion research has to deal with a fundamental dilemma. On the one hand, it needs to draw on field evidence, which is obtained in the real world in which all components of the urban system – buildings, people, traffic, weather – interact, not in a way we can control completely, but in whatever way they do interact. On the other hand, it needs to draw on wind tunnel evidence, which is controlled in a sense that allows particular processes involved in flow and dispersion to be replicated, compared and contrasted.
What this means is that field data and wind tunnel data are not substitutes for each other, but provide complementary evidence to answer different questions. Field data tells us what happens in the urban environment. Wind tunnel studies are useful for isolating physical processes which are hard to separate in full-scale field measurements.
DAPPLE’s approach had a strength in not having to choose between field data and wind tunnel data. The field campaigns in DAPPLE were used in conjunction with wind tunnel studies and numerical modelling to provide an understanding of urban processes in street and neighbourhood environments from the standpoint of air quality, traffic, buildings, and the interactions of these, and of the people in the street environment.
The Value of Field Data
Field data are valuable because they represent reality. Cities are not an abstract system with idealised inputs and simple boundary conditions. They include non-uniform building shapes, time-dependent traffic flows, fluctuating weather conditions, surface temperature, background pollution and moving populations. All these factors influence the dispersion of pollutants and people’s exposure.
Field data allow the detection of features which cannot necessarily be replicated in wind tunnel or laboratory experiments and include local flow at junctions, the effect of traffic congestion on emissions, changes in pollutant concentrations during the day, and exposure at different locations and times for individuals moving through street canyons.
For DAPPLE, these measurements were carried out in central London in the area around the Marylebone Road and Gloucester Place. The array of measurements includes street-level instrumentation, roof-level and BT Tower measurements, traffic data, releases of tracer gas from the road and the street network, and the exposure measurements made by DAPPLE volunteers in the area.
All these data together can provide information on the actual experience of a pedestrian, a cyclist or a person waiting at a bus stop in the busy and polluted streets of London, for which one measurement at a single location would not give a full picture of the variability of the conditions. Field data are more relevant to what decision makers have to deal with in practice.
The Limitations of Field Measurements
Field data are valuable because they represent reality, but this reality makes the measurements more difficult to interpret. In a real urban area, conditions are never fixed. Wind speed and wind direction change. Traffic changes. Pollution concentrations change. If one release or one measurement is made, another cannot be performed which is identical in terms of the boundary conditions and the meteorological situation.
This can make it difficult to interpret results. If a field measurement is compared with another field measurement and these differ, this may be because of a change in traffic, a change in wind direction and speed, background concentrations, or it may be because of something that is not known because the resolution was not high enough to detect it.
A numerical model may not replicate a field measurement because the model has an error or an assumption is not a good one. The input data could be missing some information. Or, more subtly, there may be real-world change that was not accounted for by the input conditions.
One aspect of this is that street-level air pollution is difficult to predict. This is because it is not only how much is emitted, but how the emitted pollution disperses across a complicated and changing urban environment.
What Wind Tunnel Data Adds
Wind tunnel modelling introduces an element of control.
When you are doing an experiment in a wind tunnel, you can place your physical urban area model under known repeatable flow conditions and control these more effectively than in the field. Wind direction, flow speed and setup can all be adjusted very precisely.
This enables you to repeat the experiment a lot and see whether results are similar in nominally identical conditions. This is helpful for studies of urban dispersion, where short-range dispersion can vary from one release event to the next, and one result does not give much understanding of the statistics of the process.
A number of repeats is better than just one, so that you can be more sure about your understanding of the results and not have one result that might be due to one set of conditions.
Wind tunnel experiments can also allow you to look at different effects that may be contributing to urban dispersion. For example, you could study how street geometry, building height, junction layout or source position influences the flow. These questions are hard to answer in the field when there are so many factors that are changing all the time.
This is why wind tunnel modelling was such an important part of the DAPPLE technical work, as it gave a way of understanding processes that occur in a controllable and repeatable way in terms of flow and dispersion processes.
The Limits of Wind Tunnel Experiments
But a wind tunnel is not a city, and a wind tunnel physical model cannot capture all the elements of an urban environment. For example, traffic flow may be approximated, thermal effects may not be included or modelled, background pollution may not be varying, and pedestrian or vehicle traffic and city activity may not be able to be captured to scale, at least in a reasonable way.
There are also questions about how to capture relevant physical processes in a scale wind tunnel model. In general, you have to use an approximate method, and that requires assumptions and understanding to determine what you need to capture in the wind tunnel, and what you need to leave out.
So, the wind tunnel data should not be interpreted as a full replacement for field measurements, but instead as a useful complementary source of information because of the control and repeatability of the wind tunnel. However, there are aspects that the wind tunnel cannot account for because of the limits in physical scale and ability to fully replicate a physical urban environment.
Ultimately, the best interpretation of wind tunnel results is that you look at this data alongside field data, rather than considering the wind tunnel results in isolation.
Why the Two Data Sources Are Complementary
It is important to understand that while field measurements show us what actually happened in a real urban environment, and are useful for that reason, these measurements are influenced by changing flow conditions. Wind tunnel measurements provide a more controlled environment, in the sense that they can be repeated under exactly the same conditions.
While wind tunnel testing permits multiple trials under standardised settings, it necessarily abstracts elements of the actual environment. These methods, when used in tandem, empower scientists to pose more refined inquiries.
Should a particular dispersion signature manifest in both wind tunnel and field measurements, one’s confidence in a given reading is bolstered. Should both sources diverge, this difference is in itself potentially instructive, revealing overlooked physical processes, flawed inputs, scale effects, or other complexities in the real world not replicated in a laboratory setting.
This is crucial for understanding the dynamics of street canyons, intersections and densely built-up districts where pollutant transport may hinge in large part on very specific local features and the wind direction. Observations on site represent the complete situation, while laboratory work indicates which features may be the major cause of particular outcomes.
Tracer Experiments Need Both Sources of Evidence
Tracer experiments offer a prime example of why we need evidence from both sources.
In the field, a tracer experiment is a way of understanding how a material disperses through street canyons in real weather and traffic conditions, highlighting potentially unforeseen transport routes, rapid short-term changes, movement down side streets, and variations between different receptor positions.
Repeating such an experiment exactly the same way in the field is often problematic since the necessary weather conditions to achieve sufficient releases may not be met before conditions change. In contrast, tracer experiments can be carried out repeatedly in wind tunnels under standardised flow conditions, and thus provide a more complete understanding of the dispersion process in statistical terms.
However, such experiments cannot reflect all the effects of real-world flows on full-scale releases. These issues were explored in greater depth in the article about what tracer experiments reveal about pollution movement in cities; here the most pertinent fact is that tracer investigations are more robust when field and wind tunnel results are combined.
Model Reliability Depends on Evidence Quality
The reliability of urban dispersion models rests on the quality of evidence on which they are based.
Dispersion models estimate pollutant transport in the absence of measured concentration data and are relied on in assessment of air quality, strategic planning, emergency response and exposure determination. But we can only use such models if we have good reason to believe they will perform well.
Field and wind tunnel experiments enable us to determine that this is the case.
Field data enable a model to be judged against the full complexity of an urban site by comparing the predicted concentrations and plume movements with measured concentrations in streets and the influence of meteorology on urban flows.
Laboratory data enable a model to be tested by repeating comparison of the predicted concentrations and plume movements under controlled conditions to establish whether a model captures key flow and dispersion processes.
The DAPPLE-HO project placed particular emphasis on short-range dispersion and repeat experiments and on evaluation of urban dispersion models. This is in recognition of the fact that policy-related problems, such as the modelling used in planning or for emergency responses, cannot be relied upon if based only on assumptions, but require evidence that the models perform well.
Implications for Policy and Planning
Why does this field-versus-wind-tunnel question matter for policy, and why does a combined evidence base matter for the way urban air quality policies are made?
This matters because many urban air quality decisions involve models and measurements. A local authority may want to know how traffic changes can affect exposure; emergency planners may want to know how much a short-term release can travel; or urban designers may want to know whether new street arrangements will improve or worsen ventilation.
In each case, the quality of the evidence will help determine the reliability of the decision.
We know that field data alone may not be variable enough to allow a good explanation of the process, and that wind tunnel data alone may be too idealised to allow an accurate description of a complex problem. But we also know that if we have both, it will help to have data showing not only how things are, but also what the reasons might be.
Furthermore, this can contribute to a better understanding of how the available monitoring data are best used in the decision-making process. It is unlikely that the data from a single fixed monitor will be representative of the concentrations from all the nearby streets.
Field campaigns can be used to explore this variability in monitoring data, while wind tunnel results can help to understand the reasons for this variability.
Physical Modelling and Digital Modelling
There is also an important relation between wind tunnel data and physical and digital modelling in general:
- Computational models can test a large number of scenarios quickly, but they are still based upon models and evidence that require development and evaluation.
- Physical modelling can provide controlled data that can be used to test how well numerical models reproduce realistic flow and dispersion.
- A broader view of this is illustrated by PHYSMOD 2013, which focused on the physical modelling of atmospheric flow and dispersion, including urban dispersion, building effects, boundary layer modelling and validation of numerical tools.
The conclusion is not that one or the other approach is needed, but that all the elements — field, wind tunnel and digital — can make important contributions to urban dispersion research if the elements are considered part of the same evidence system.
A Combined Evidence Base
Both field data and wind tunnel data have disadvantages. In particular, field data allow realistic, but uncontrolled, measurements. Wind tunnel data allow controlled, but necessarily idealised, measurements.
For research on urban dispersion, however, there is a need for both. This is because the real environment of the city is both complex in terms of the physical problem and important in terms of policy needs. It is not sufficient to understand dispersion, and to make the right policy decisions if one understands it only in ideal circumstances, but one is unable to make sense of real data in complex circumstances.
DAPPLE’s contribution to urban dispersion was to bring all the strands together: field campaigns, tracer experiments, exposure measurements, wind tunnel studies and numerical modelling. This kind of combined approach remains necessary to understand how pollutants travel within the city.
The policy implication for urban air quality is simple. It is important to make good decisions, but good decisions require good evidence, and better evidence will be more able to describe the real city, explore the reasons for the processes that are being observed, and evaluate the models that are used to predict what will happen in the future.
Field data and wind tunnel data do different things, but urban dispersion research will be better if it uses both.
