Tel: 01904 321218
My research is focussed on developing a better understanding of urban air pollution. One of the most important findings was the discovery that emissions of nitrogen dioxide (NO2) were increasing (Carslaw, 2005). Before that work, directly emitted NO2 from vehicles was largely ignored and unquantified. Since that time, urban NO2 concentrations in many locations across Europe have become important because they are in excess of European air quality limits. Some of the highest ambient measurements of urban NO2 are controlled by directly emitted NO2 from road vehicles. Much of my work since that time has aimed to better quantify emissions vehicle emissions under urban driving conditions.
Another area of focus is the development of receptor modelling approaches to improve understanding of the influence of different emission sources on ambient concentrations. A component of this work includes the development of an R package called openair. The package is widely used throughout the world by academia the public and private sectors.
The openair project started in 2008 with 3-year funding from the Natural Environment Research Council (NERC). The principal aim is to make available innovative data analysis techniques for air pollution data – or more generally atmospheric composition data. openair is an R package that is openly and freely available to anyone. Some of the capabilities of openair include:
(see also the publications under receptor modelling)
Understanding sources of air pollution is essential if air pollution is to be effectively controlled. There are numerous ways of using measurement data itself to draw conclusions about the nature of different emission sources under the general area of receptor modelling. Our recent interests have focussed on developing an approach based on bivariate polar plots. These plots provide an effective way in which to better understand and isolate different sources of air pollution.
As an example of the techniques developed, the plots below show the analysis of data from an air pollution monitoring site in Scunthorpe Town. To the east of the site is a highly complex integrated steelworks. Approximately 30 km from the site are large point sources.
A common approach when considering such data is to plot a pollution rose, similar to those shown in the Figure below (a) and (c) for NOx and SO2, respectively. These types of plots provide information on the direction of principal sources – but little else. However, by bringing in a third variable (wind speed in this case), a surface can be plotted as shown in (b) and (d). In these plots zero wind speed is shown at the centre and higher wind speeds increase radially outwards. The colour represents the concentration. In the case of NOx (b) it can be seen that the highest concentrations occur under the lowest wind speed conditions, with some indication of higher concentrations to the south-east. Such behaviour is typical of ground-level non buoyant sources such as road traffic emissions and is often seen at background sites in urban areas.
By contrast, the plot for SO2 (d) is very different to that of NOx i.e. the highest concentrations occur under the higher wind speeds. Such behaviour is indicative of buoyant emissions from a chimney stack. The actual behaviour of different pollutants depends on many factors but differing behaviour can help provide information on the nature of the emission source rather than simply showing the direction in which it is located. Note the radial axis does not need to be wind speed; it could be any numeric variable. The key issue is that the variable helps to differentiate different source types e.g. some other measure of atmospheric stability.
In more recent work (Uria-Tellaetxe and Carslaw, 2014) we have developed the bivariate polar plot approach further. Plots (b) and (d) above show the mean concentrations. There is however, a distribution of concentrations and it can be shown that considering the distribution specifically can help further differentiate source types. The approach is based on considering the joint wind speed / direction probability that a certain range of concentrations exists. By ‘scanning’ a range of concentrations, it is possible to reveal conditions where specific sources have their greatest impact. The plots below show how different major sources become prominent at certain concentration intervals. The paper contains more detailed information on the technique.
Our work on vehicle emissions has focussed on developing a much better understanding of the emissions of NOx and NO2 from individual vehicles. Much of the new insight has come from the use of a vehicle emission remote sensing technique developed at the University of Denver. Importantly, the instrument from the University of Denver measures NO and NO2 (commercial instruments only measure NO, which is a major drawback when measuring the emissions from modern diesel vehicles).
The Figure below shows how the thousands of measurements used for ‘Type Approval’ of vehicle emissions in Europe compares with on-road remote sensing measurements from our campaigns for diesel passenger cars. The results show that post 2005 vehicles typically emit about four times as much NOx under real driving conditions compared with when they are driven over the Type Approval test cycle. Individual vehicle manufacturers and vehicle models will of course be lower and higher than shown.
The Figure below shows the clear increase in directly emitted NO2 emissions from diesel vehicles in recent years. Direct emissions of NO2 from petrol-based vehicles are extremely low. An important characteristic of many modern diesel vehicles is that they emit substantially higher emissions of total NOx compared with petrol vehicles – and the proportion of that NOx that is NO2 is much higher. For urban air pollution, this characteristic of diesel vehicles is at the heart of the NO2 air pollution issues across Europe.