Analysis of station classification and network design
INERIS (Laure Malherbe, Anthony Ung), NILU (Philipp Schneider), RIVM (Frank de Leeuw, Benno Jimmink)
18th EIONET meeting -Dublin- 25th October 2013
Context
An increasing amount of available data on air quality in Europe
Extension of data coverage for all pollutants both in time and space Much effort dedicated to data QA/QC (e.g. JRC-AQUILA quality
programmes) Information on siting requested in current and future reporting
obligations
18th EIONET meeting, Dublin, 25th October 2013
O3 and PM10 monitoring stations for which data have been reported to AirBase
AirBase v7, www.eea.europa.eu
(JRC- AQUILA Position Paper on siting criteria and station classification)
Context
...but still limited information on the monitoring strategies underlying site selection; on the fitness for purpose of the selected measurement
locations.
An encouragement to refine existing station classification schemes or develop
supplementary ones, to develop meta-information describing the station surroundings
(land use, population density,...)
18th EIONET meeting, Dublin, 25th October 2013
(JRC- AQUILA Position Paper on siting criteria and station classification)
Objectives of the study
First part : evaluation of the network design from several angles: evolution from 1996 fulfilment of the EUROAIRNET criteria compliance with the AQD
Second part: a supplementary classification scheme (presented by L. Rouïl at the17th EIONET meeting, 2012): update the classification according to Joly and Peuch (2012)
methodology and check its robustness analyse the results on the European scale investigate specific situations
NB: study mainly focused on NO2, O3 and PM
18th EIONET meeting, Dublin, 25th October 2013
Evaluation of the network
18th EIONET meeting, Dublin, 25th October 2013
Evolution of the network
Selected years: 1996: state before the implementation of the Framework AQ
Directive (information for the majority of EU15 Members) 2004: from EU15 to EU 25 2007: from EU25 to EU27 2011: the most recent year available in AirBase
18th EIONET meeting, Dublin, 25th October 2013
class type of area type of station
(sub)urban background (U)
urban backgroundsuburban background
traffic (T)
urban trafficsuburban trafficrural trafficunknown traffic
regional background (R)
rural background
industrial (I)
urban industrialsuburban industrialrural industrialunknown industrial
Considered station categories:
18th EIONET meeting, Dublin, 25th October 2013
1996 2007
2004 2011
PM10
Monitoring criteria
18th EIONET meeting, Dublin, 25th October 2013
EURO-AIRNET (Larssen et al., 1999): number of cities to be included in a European
representative network:• all large cities (>500,000)• 25% of medium cities (250,000-500,000)• 10% of small cities (50,000-250,000)
Agg/zone >UAT L-UAT
250 2 1
500 3 2
750 3 2
1000 4 2
1500 6 3
2000 7 3
2750 8 4
3750 10 4
4750 11 6
6000 13 6
more 15 7
Minimum requirements for PM (PM10 + PM2.5)
AQ Directive The number of stations in an
agglomeration/zone depends on population and current AQ status
NB: Countries may also use modelling as a supplementary assessment tool ; in that case these numbers may be reduced by up to 50% under the conditions set in Dir. 2008/50/CE, Art. 7
Monitoring criteria
18th EIONET meeting, Dublin, 25th October 2013
Application of EUROAIRNET criteria:
Minimum monitoring coverageActual PM10
monitoring coverage
Actual NO2
monitoring coverage
all large cities (>500,000) 72 of 73 (99%) 59 of 73 (81%)
25% of medium cities (250,000-500,000)
93 of 116 (80%)
92 of 116 (79%)
10% of small cities (50,000-250,000)
555 of 750 (74%)
528 of 750 (70%)
Monitoring criteria
18th EIONET meeting, Dublin, 25th October 2013
Application of AQD criteria:Ex: PM monitoring, 702 zones/agglomerations common to 1996, 2004, 2007 and 2011. Compliance with AQD considering the assessment regimes applicable in 2011:
Results for NO2 monitoring:
Number of zones for which:
1996 2004 2007 2011
Nstations = Nmin 32 108 102 113
Nstations > Nmin 6 193 306 423
Nstations < Nmin 664 401 294 166
Number of zones for which:
1996 2004 2007 2011
Nstations = Nmin 211 174 132 105
Nstations > Nmin 15 337 462 516
Nstations < Nmin 476 191 108 81
Supplementary classification according to Joly & Peuch (2012)
methodology
18th EIONET meeting, Dublin, 25th October 2013
Brief recall of the methodology
Methodology developed in the framework of GMES/MACC program
Objective: establishing a pollutant-specific objective classification of stations based on the temporal variability of the observation data
For each considered pollutant, time series are summarized by eight indicators characterizing the diurnal cycle, the weekend effect and the high frequency variations.
The classification is performed in two stages:1) Definition of the classes, from class1 to 10, with a selected set of
stations. A linear discriminant analysis is performed so as to best discriminate between rural stations and stations most influenced by human activities (“urban + traffic” sites).
2) Classification a posteriori of the other stations.
ETC/ACM Technical Paper 2012/17 (2013)
18th EIONET meeting, Dublin, 25th October 2013
Update and analysis with AirBase v7
18th EIONET meeting, Dublin, 25th October 2013
Spatial distribution of the classes
O3
PM10
NO2
In some countries, all station classes are present. For some pollutants, other countries only have low or high station classes. Classification is missing for some stations (criteria not filled, missing data, only daily values reported)
More stations have been classified:
18th EIONET meeting, Dublin, 25th October 2013
Number of stations classified in ETC/ACM (2013) study(AirBase v6, 2002-2010)
Number of stations classified in this study(AirBase v7, 2002-2011)
NO2 2697 3136
PM10 1822 2248
O3 2098 2349
Monitoring stations not classified last year but classified in this study
Update and analysis with AirBase v7
PM10
O3
NO2
As in 2012, the classification has been analysed in relation with EoI classification and auxiliary variables (population density, land use) This analysis confirms the robustness of the methodology.
18th EIONET meeting, Dublin, 25th October 2013
Distribution of PM10 classes as a function of EoI classification
Mean and median population density around PM10 measurement
stations for each class
Update and analysis with AirBase v7
Different types of specific situations have been identified : stations for which the classification (according to Joly & Peuch, 2012)
does not well match the types of area and site provided in AirBase; stations in specific environments (e.g.: high population density); stations displaying very different classes according to the measured
pollutant.
18th EIONET meeting, Dublin, 25th October 2013
Interesting cases
18th EIONET meeting, Dublin, 25th October 2013
Distribution of population density in each PM10 class
Three stations have a population density higher than 20000 inhab./km2
They are located in the same area: Paris
They are all urban background but do not have the same class.
Interesting cases: example
18th EIONET meeting, Dublin, 25th October 2013
PM10 classes in Berlin
Berlin agglomeration contains all EoI types of area/station.It also contains almost all the class numbers, with growing numbers from rural background to traffic sites.As in Paris, two urban background stations are interesting for study.
Interesting cases: example
General conclusions
Station classification and detailed description of the station surroundings provide helpful support: to interpret air quality data to have a better idea of the station representativeness to select the most relevant sets of stations for trend analysis,
model evaluation, data assimilation, air quality mapping, impact studies...
This can be achieved by the joint use of different classification schemes such as EoI (type of area/type of
site) and Joly & Peuch (2012) methodology auxiliary data such as population density, land cover, emissions...
It is proposed to compile and make such information available to data users.
18th EIONET meeting, Dublin, 25th October 2013
Proposed content of the spreadsheet
Station code Coordinates Ozone classification (if applicable) EoI Classification:
Type of area Type of station Characteristics of zone (if available)
Classification according to Joly & Peuch methodology (2012): Class number for O3, NO2 and PM10
Dominant emission sector(s) (if available) LAU and NUTS codes (EBM) Population density within 1 km radius (JRC database and ORNL) Proportion of each main land cover class within 1 km radius
(CORINE) Other remarks
18th EIONET meeting, Dublin, 25th October 2013
A premilinary Excel file is available => link