• No results found

We conclude this report with a summary of suggestions on where to direct future research work on spatial representativeness:

The assessment of the spatial representativeness of air quality monitoring stations remains an outstanding issue that has substantial links to several highly topical areas, including risk assessment and population exposure (ref: Directive 2008/50/EC and Implementing Decision 2011/850/EU), the design of monitoring networks, model development, model evaluation and data assimilation. . Even if different purposes of estimating SR might cause some conflict of goals, there is a substantial room for improvements towards a higher transparency and the need for a better comparability of SR estimates.

11.1 Proposed work with regards to harmonisation

The conclusions that can be drawn from the IE and from this report underline the need for a more harmonised definition of the concept of “the area of representativeness”

and for consistent and transparent criteria used for its quantification. We have pointed out that this objective will first require establishing a common framework for

1. SR definitions and 2. SR similarity criteria

3. harmonising the related terminologies

For the aim of harmonisation, however, the concept of SR probably requires a more fundamental paradigm shift in its definition, which has been detailed in the previous chapter 10. In that chapter, we proposed a modular approach towards better SR characterisation, in which a more clear distinction needs to be made between the four different aspects of spatial representativeness, repeated here:

1. The purpose of evaluating SR in a specific case of application

2. The set of SR metrics / SR characteristics required for this purpose 3. Context related definitions of SR metrics

4. The technical methods for estimating a particular SR metric

11.2 Proposed work regarding methodological evaluations

As explained in the previous chapters, the IE concerned the full user manipulable parameter space, including (i) the choice of a subset from the data available, (ii) the SR method as such, (iii) the similarity criteria definitions, and (iv) the parameter values chosen for these similarity criteria. Unfortunately, current outputs do in turn not enable us to distinguish between the individual influences of these different groups of parameters.

We thus consider that a range of statistical analyses would be important to be conducted before getting into serious discussion about harmonisation of the selection of parameter values for the similarity criteria and SR methods. Amongst these suggestions are:

● Quantitative sensitivity analysis (permutations, Monte Carlo simulations) to investigate how the parameterisation of the similarity criteria / threshold values influences the estimation of SR areas. These tasks have been laid out in more detail in chapter 10.5.

● How can SR codes be inverted to find optimal station positions?

● Investigate how the SR area evolves over different evaluation periods (i.e. time series of daily SR area estimates over a year, time series of SR for the hours of the day – i.e. 24 averages of 365 hours, etc.)

More detailed information about these suggestions have been outlined and can be found in chapter 10.

11.3 Proposed work regarding measurements

It has been pointed out (chapter 10) that an ultimate validation of SR methodologies can never be done on purely synthetic data or on modelling, but will require a strong consideration of air quality measurement with a high spatial resolution. On the one hand, a future link can thus be seen between the assessment of spatial representativeness and the emerging research field of (low-cost) mobile air quality sensors. On the other hand SR could be compared with pollution distribution estimated using – if available – measurements obtained by high spatial resolution sampling campaigns with diffusive samplers

List of Abbreviations and Definitions

AQD air quality Directive

AQMS air quality monitoring station or air quality monitoring site AQUILA Network of Air Quality Reference Laboratories

CFD computational fluid dynamics

FAIRMODE Forum for Air Quality Modelling in Europe PCA principal component analyses

SR spatial representativeness IE intercomparison exercise SR area spatial representativeness area

List of Figures

Figure 1. Overview of the annual average concentration fields obtained for PM10, NO2 and O3 for the modelling year 2012. ... 10  Figure 2. Time series of spatial mean, spatial standard deviation, and relative spatial standard deviation of virtual monitoring points. ... 13  Figure 3. Examples of SR area estimates obtained for NO2 at the urban-background site Antwerpen-Linkeroever (site v7). ... 24  Figure 4. Examples of SR area estimates obtained for O3 at the urban-background site Schoten (site v17). ... 25  Figure 5. Examples of SR area estimates obtained for PM10 at the traffic site Borgerhout-Straatkant (site v216)... 25  Figure 6. Summary chart of spatial representativeness areas (SR area in km2) obtained for the pollutants NO2, O3 and PM10 at the urban-background sites

Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site Borgerhout-Straatkant (v216).

... 27  Figure 7. Spatial representativeness area estimates (SR area in km2) obtained for the pollutant NO2 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site Borgerhout-Straatkant (v216). ... 28  Figure 8. Spatial representativeness area estimates (SR area in km2) obtained for the pollutant O3 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17). ... 29  Figure 9. Spatial representativeness area estimates (SR area in km2) obtained for the pollutant PM10 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site Borgerhout-Straatkant (v216). ... 30  Figure 10. Number of inhabitants within the estimated areas of representativeness (population in thousands) obtained for the pollutant NO2 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site

Borgerhout-Straatkant (v216). ... 32  Figure 11. Number of inhabitants within the estimated areas of representativeness (population in thousands) obtained for the pollutant O3 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17). ... 33  Figure 12. Number of inhabitants within the estimated areas of representativeness (population in thousands) obtained for the pollutant PM10 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site

Borgerhout-Straatkant (v216). ... 34  Figure 13. Example of the MLA indicator calculated between ENEA and EPAIE for O3 at the urban-background site Schoten (site v17). ... 36  Figure 14. MLA indicators (pairwise level of agreement) of the SR estimates obtained for NO2 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site Borgerhout-Straatkant (v216). ... 38  Figure 15. MLA indicators (pairwise level of agreement) of the SR estimates obtained for O3 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17). ... 39  Figure 16. MLA indicators (pairwise level of agreement) of the SR estimates obtained for PM10 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site Borgerhout-Straatkant (v216). ... 40  Figure 17. Lumped MLA indicators obtained by combining MLAs for all three pollutants (NO2, O3 and PM10) at all three sites v7, v17 and v216 (top panel ), and at the two

background sites v7 and v17 (bottom panel), respectively. ... 42 

List of Tables

Table 1. Summary statistics of the time series of 341 virtual monitoring points ... 11  Table 2. List of participating teams and institutions ... 14  Table 3. Overview of input data used by the different teams. Grey background indicates data additional to the shared Antwerp dataset. ... 19  Table 4. Overview of input data used by the different teams for traffic sites. Grey background indicates data additional to the shared Antwerp dataset. ... 20  Table 5. Overview of input data used by the different teams for background sites.

Grey background indicates data additional to the shared Antwerp dataset. ... 21  Table 6. Overview of results received from the different teams ... 23  Table 7. Overview of spatial representativeness estimates (SR area in km2) obtained for the pollutants NO2, O3 and PM10 at the urban-background sites Antwerpen-Linkeroever (v7) and Schoten (v17), and at the traffic site Borgerhout-Straatkant (v216). ... 27  Table 8. Results of the intersection analysis: areas of the total union and the final

intersection of all SR area estimates. ... 35 

Annexes