3.1. Spatial Density Analysis
To characterise the spatial density of the Smogmobile NO2 measurements, ArcGIS was used to define two different grid resolutions (100 m and 200 m grid square networks) within an ‘Area of Interest’ (grey area in Figures below) covering the City of London and immediate environs. The average NO2 concentration measured by the Smogmobile was adopted for the entire road line between two consecutive points. Each line between two points could overlap with one or more grid square; hence, depending on the grid resolution adopted, the average NO2 concentration was assigned to each of the grid/s square/s, and then an average was calculated among all values (or Smogmobile data points) falling within that grid square. In this study, that was the best approximation possible; however, in a subsequent study, the data measurements will be performed at a higher resolution (sec-by-sec) eliminating the problem of each measurement representing an average concentration among multiple grid squares.
The output of the density plotting process is presented below in Figure 3
and Figure 4
for 100 m and 200 m grid square networks, respectively. White areas within the City of London boundary indicate that there were no Smogmobile data for that location/grid square. Numbers within each grid cell indicate the number of samples used to calculate the average pollution concentration levels. This analysis indicates that average NO2 concentrations measured by the Smogmobile were generally higher in the south of the City, where high volume traffic and the presence of busier junctions (roundabouts and signalised) produce higher emissions leading, for the most part, to values higher than 40 µg/m3
(the annual limit value) across the extent of the City.
Further, some of the highest concentrations were recorded inside the road tunnel in Upper Thames St. (A3211) due to scarce ventilation.
Both Figure 3
and Figure 4
indicate that only an area covering approximately six 200 m × 200 m grid squares (or twenty-four 100 m × 100 m grid squares) within the City boundary experienced average NO2
concentrations of less than or equal to 40 µg/m3
. This is equivalent to an area of approximately 0.24 km2
, 7.62% of the entire area within the CoL boundary (3.15 km2
3.2. Time Series Analysis
presents, at 15-minute intervals (Cumulative average is calculated starting from 1 January; when adding a new record, the average is re-calculated based on all of the measurements/data points.), cumulative average NO2
concentrations in 2017 at CT3 (Sir John Cass School), CT4 (Beech Street), and CT6 (Walbrook Wharf), along with corresponding monthly average NO2
concentrations from the diffusion tubes (expressed as the unadjusted average of all NO2
diffusion tube measurements in each month).
The CMS data indicate that (following an initial period of measurement instability at the start of the calendar year) after between four and six months of monitoring activity, NO2 concentrations stabilise to approximately match the final annual mean concentration at the end of the year.
The final average NO2 concentration of all three CMS (70 µg/m3) is closer to the final annual mean NO2 concentration derived from the monthly diffusion tube measurements (57 µg/m3, unadjusted) than any of the individually derived annual average values for CT3, CT4, and CT6, and is within ±25% (according to DEFRA’s Technical Guidance) LAQM.TG(09) states that diffusion tube measurement uncertainty is often cited as being ±25% of the final annual mean diffusion tube concentration (57 µg/m3, 55 µg/m3 bias adjusted).
below presents the same annual cumulative average as per Figure 5
, but includes also the cumulative average NO2
concentration for each CMS when considering the date and time matching the available Smogmobile data (labelled ‘week_cumul.avg’ later in Figure 8) during the survey period in March 2017. This means that several overnight periods and some periods when the Smogmobile was not monitoring or was not on-street have been excluded from the average.
CT3 and CT4 show a reasonable correlation between the 5-day and annual average, while CT6 (132 µg/m3) and all CMS (86 µg/m3) are less well correlated with only CT6 falling outside ±25% of the corresponding annual mean values (93 µg/m3 and 70 µg/m3, respectively).
shows the cumulative average for each of the three CMS (and the average of all three CMS) during the entire Smogmobile Survey period (from 7:30 a.m. on the 23 March 2017 until 13:00 on the 29 March 2017). Average values for CT3, CT4, CT6, and all CTs are 38 µg/m3
, 82 µg/m3
, 116 µg/m3
and 78 µg/m3
, respectively, consistently higher than those in Figure 9, between 5% and 12%.
This provides evidence that it is really important to design and plan in advance the time and location where a mobile monitoring system should operate to avoid observing or measuring concentration and in particular obtain averages that may not be representative of the real situation and the annual mean. In this case, excluding overnight observations has the impact to return averages that are generally higher (5 to 12%) than those normally observed by continuous monitoring systems located on the street.
and Figure 9
present, for the five-day Smogmobile survey period, cumulative average NO2
concentrations for all three CMS (expressed as the cumulative average of the NO2
concentrations measured at CT3, CT4, and CT6), all diffusion tubes (expressed as the average NO2
concentration measured from all diffusion tubes in March 2017), and the Smogmobile (expressed as the cumulative average of all NO2
concentrations measured by the Smogmobile) without stationary car park measurements (Figure 9
) and with all measurements (Figure 8
In Figure 8
, the cumulative average CMS and Smogmobile concentrations tend to stabilise and converge, at the beginning of 28 March 2017, close to a value of approximately 86 µg/m3
. This initial result indicates that the Smogmobile is capable of representing the average concentration across the City area, assuming that the CMS are considered to be broadly representative of air quality levels across the City.
However, when looking at Figure 9
, where the stationary carpark measurements have been excluded from the analysis, the results are greatly different. Almost immediately, after few hours of mobile monitoring, the cumulative average Smogmobile NO2
concentration tends to stabilise around a value of 60 µg/m3
, which is the same average value obtained using the concentrations reported for March 2017 by 48 diffusion tubes (59.5 µg/m3
) spread across the CoL area.
The analysis performed and its results pose some interesting questions in terms of the most appropriate duration of a mobile campaign and what periods of the year should be targeted, but fundamentally it opens up the possibility for the Smogmobile approach to be a potential alternative to, or to complement, traditional static methods (such as CMS and DT). However, the limited duration of the Smogmobile monitoring campaign in this study presents a serious limitation in terms of the accuracy of the statistical results; hence, a longer monitoring campaign, possibly split between two or even three periods of the year, would be advisable to capture both seasonal variations and to give more statistical robustness to the results and findings.
3.3. Comparative Analysis Smogmobile, CMS, and DTs
Based on the density plot analysis presented in Section 3.1
and considering the 200 m cell/grid network identified, by taking the three cells where the CMS are located and using the same ID reference to obtain the average NO2
concentration of the Smogmobile data points and diffusion tubes falling within those cells, a comparative analysis between Smogmobile, CMS, and diffusion tube average concentrations was carried out (see Figure 10
As can be observed, the Smogmobile average NO2 concentration is generally closer to the CMS than the diffusion tubes. For CT3 (Sir John Cass School), which is classified as an urban background site, the average NO2 concentration of the Smogmobile is 10 µg/m3 lower than that of the CMS and within a ±25% error, whilst the diffusion tube average concentration is 8 µg/m3 higher than that of the CMS.
For CT4, the Smogmobile average NO2 concentration is much closer (16% lower) to the CMS average than the diffusion tube average (44% lower). For CT6, the average Smogmobile NO2 concentration is very close (5% lower) to the CMS average.
These results provide further confidence that mobile monitoring is a suitable alternative, or can complement, static sensors. However, it is important to stress the difference in the time scale of the average values used/presented. The diffusion tube values are based on a one-month period (March 2017) whilst for the Smogmobile and the CMS, measurement data are available for a five-day period only. A truly comparative study will require, again, a longer duration monitoring campaign.