JRC Initiative on Source Apportionment Harmonization in Europe: Outline and Developments

Claudio A. BelisJoint Research Centre – IES – Air and Climate Unit

FAIRMODE MEETING Antwerp 10th-12th April 2013

• Structure of the presentation

• JRC initative on Source Apportionment • Progress of work:

• Survey and review of RM studies in Europe• Intercomparison for RM first step• Intercomparison for RM second step• RM Technical Protocol

Receptor Models• Use measured concentrations at the receptor• Are based on least square multivariate analysis to solve a

chemical mass balance linear equation• Mainly used for PM, PAH and VOC• Not appropriate for reactive species• Appropriate for urban and regional scales

Basic equationij

P

p

jpipij efgx 1

Concentration of the jth species in the pth source

Contribution of pth source to ith sample

Concentration of the jth

species in the ith

sample

JRC INITIATIVE ON SOURCE

APPORTIONMENT HARMONIZATION

INTERCOMPARISON EXERCISE FOR RM

assess model performances and

quantify uncertainty

CRITICAL REVIEW AND META-ANALYSIS OF SA

STUDIES

take advantage of available data to

estimate sources at the European level

COMMON RECEPTOR MODELLING TECHNICAL

PROTOCOL

find common procedures and criteria

to assure quality standards and improve comparability among

studies

FIRST REVIEW ON RM IN EUROPE

assess the impact of the metodology, list used

tools and identify needs

FAIRMODE WG1 SG 2 ONNATURAL SOURCES AND

SOURCE APPORTIONMENT

CompletedKaragulian & Belis, 2012 IJEP

CompletedBelis et al. 2013 AE

First step completedKaragulian & Belis, 2012 EUR REPORT

Second step: in progress

In progress

APPRAISAL WP2Receptor and source

oriented models

Critical Review of RM methods and quantification of main PM

Sources in Europe

- Critical discussion of methods used in Europe- Meta-analysis of 272 records present in more than 100 papers and reports published until the beginning of 2012:- Identification of main source categories.- Description of geographical and seasonal variation of these sources were studied and mapped.- A special analysis of PM concentrations was made to assess the causes of exceedances

Critical Review of RM methods and quantification of PM Sources in Europe

Type Examples

Exploratory methods Enrichment factor, tracer method,

incremental approach

Chemical mass balance EPA CMB 8.2

Eigenvector based models PCA, UNMIX

Factor analysis without

constraints

FA, APCFA

Positive matrix factorization PMF2, EPA PMF v3

Hybrid trajectory based

models

Analysis of wind direction

Analysis of backward trajectories

Hybrid expanded models PMF solved with ME-2, COPREM

Classification of receptor models with particular reference to those used in Europe

Critical Review of RM methods and quantification of PM Sources in Europe

ANALYSIS OF WIND DIRECTION

Conditional probability function (CPF):

Non parametric wind regression (NWR)

Pseudo deterministic receptor model (PDRM)

ANALYSIS OF BACKWARD TRAJECTORIES

Trajectory sector analysis (TSA)

Potential source contribution function (PSCF)

Simplified quantitative transport bias analysis

(SQTBA)

Trajectory mass balance (TRMB) or trmb

regression (TMBR)

Source Contributions estimated with receptor models using measured PM10 and PM2.5 chemical composition in urban background areas (preliminary results). 108 studies until 2012 - 272 records

Marker Median ± 25P-75P (box) ± Min-Max (whiskers)

SOURCES OF PARTICULATE MATTER IN EUROPE

@

six major source categories for PM were defined that comprise the majority of the individual sources apportioned in Europe

salt sulphate nitrate crustal traffic point bio. burn

SIA

Critical Review of RM methods and quantification of PM Sources in Europe

Additional source coal combustion observed in specific areas e.g. Poland

Critical Review of RM methods and quantification of PM Sources in Europe

Statistical analysis of the influence of:size fractionsite typeseasonal trendsgeographical patterns

For the OC fraction three main source categories were identified (60 records)

Critical Review of RM methods and quantification of PM Sources in Europe

Marker Median ± 25P-75P (box) ± Min-Max (whiskers)

SOURCES OF PARTICULATE MATTER IN EUROPE

PM10 SOU R C ES AN D EXC EED AN C ES

Sa lt SO4 N O3 SIA C rus ta l Tra ffic Po in t B io . bu rn .0

10

20

30

40

50

60

70

80

90

100

SC

E (

%)

PM_MASS: <= 40PM_MASS: > 40

P M 2 .5 S O URCE S A ND E X CE E DA NCE S

S a l t S O 4 NO 3 S IA Cru sta l T ra ff i c P o in t B io . b u rn .0

1 0

2 0

3 0

4 0

5 0

6 0

7 0

8 0

9 0

1 0 0

SC

E (

%)

P M _ M A S S : <= 2 5P M _ M A S S : > 2 5

A special analysis of PM concentrations that exceed the current European air quality limits indicated SIA and traffic as the most important source categories to target for abatement throughout the year together with biomass burning during the cold season.

Critical Review of RM methods and quantification of PM Sources in Europe

FINAL REMARKS• - Receptor models evolve towards tools with refined uncertainty

treatment.

• - Positive Matrix Factorization and Chemical Mass Balance are the most used models.

• - Gas-to-particle conversion is the main PM mass and particulate organic carbon source.

• - To abate exceedances, sources of secondary inorganic and traffic are the main target.

• - More long term speciated PM datasets would foster source identification studies

• - There is a need for harmonization in the methodology, with particular reference to the nomenclature of source categories.

Critical Review of RM methods and quantification of PM Sources in Europe

Intercomparison for Receptor Models

FINGERPRINTSPearson and weighted difference

SPECIES CONTRIBUTIONS (%)Pearson

Overall test sequence to identify the correspondence betweenparticipants factor/sources to each source category

If 5 out of 10 tests are nor meetthen

factor is considered dubious

Z-score(SCE)

Participant performance

Model performance

TIME TRENDSPearson

Comparison factor/sources among each otherand with reference source

Preliminary tests

Final tests

Karagulian & Belis, 2012 IJEP, 50

Final test (ISO 13528)

p

Xxscore(SCE)- z i

indicator:

SCE of participant’s source profile are optimal if:

considered coherent and satisfactory if:

“OK”

2z1 3z2 results are considered questionable if: “warning”

results are unsatisfactory if: 3z “action”

“acceptable”

1z

ORGANIZATION COUNTRY IDAEA CSIC SPAIN

Univ. Aahrus DENMARK

University of Genoa ITALYFinnish Meteorological Institute FINLAND

INERIS/LSCE FRANCE

University of Birmingham UNITED KINGDOMNorwegian Institute for Air Research

(NILU)NORWAY

Department of Physics University of Florence

ITALY

University of Milan Bicocca ITALYC.N.R. Institute for Atmospheric

Pollution ResearchITALY

IUTA e.V. GERMANYNCSR Demokritos, Environmental

Research LaboratoryGREECE

Dept. of Physics - University of Milan ITALYPaul Scherrer Institut Laboratory of

Atmospheric ChemistrySWITZERLAND

C.N.R - I.S.A.C. ITALYJOINT RESEARCH CENTRE UE

Intercomparison first step (real-world database)16 participants

First step:

Real-world database data elaboration completed

Report completed

Presentation in EAC 2012

PARTICIPANT’s Z(SCE)

unsatisfactory

acceptable

warning

OKA1 B1 B2 B3 B4 C D E F G1 G2 H I J K L M N1 N2 N3 Q S

participant code

0

1

2

3

4

5

6

7

8

9

10

ABS

(Z s

core

)

Med ian 25% -75% Min -Max

selected factors(prelim inary tests applied)

MODELS’ Z(SCE)

unsatisfactory

acceptable

warning

OK

APCFA M E2 PM F3 PM F2 PCA COPREM CM B

MODEL

0

1

2

3

4

5

6

7

8

9

10

ABS

(Z s

core

)

Med ian 25% -75% Min -Max

selected factors(prelim inary tests applied)

6 6

69

78

14

57

1) There is a good quantitative agreement between SCE. 86% of the factors meet the acceptability criteria (OK or acceptable).

2) The participant median bias in the SCEs is consistent with the 50% standard uncertainty acceptability criterion used in this evaluation.

3) PMF and CMB present the best performances, PCA showed the poorest performance but still acceptable

Intercomparison first step: results

Intercomparison second step (synthetic database)22 participants

second step:

synthetic database data elaboration in progress

Presentation in EAC 2013(planned)

ORGANIZATION COUNTRY IDAEA CSIC SPAIN

Univ. Aahrus DENMARKUniversity of Genoa ITALY

Finnish Meteorological Institute FINLANDUniversity College Cork IRELAND

University of Birmingham UNITED KINGDOMUniversity of Florence Department of Physics ITALY

Faculty of Science Charles University in Prague CZECH REPUBLICNational Institute of Public Healt and the

Environment (RIVM) THE NETHERLANDS

C.N.R. Institute for Atmospheric Pollution Research ITALYMiguel Hernández University SPAIN

NCSR Demokritos, Environmental Research Laboratory GREECE

University of Milan Dept. of Physics ITALYPaul Scherrer Institute - Laboratory of Atmospheric

Chemistry SWITZERLAND

C.N.R - I.S.A.C. ITALYAristotle University of Thessaloniki GREECE

University of Milan Bicocca ITALYUniversity of Aahrus DENMARKUniversity of Lisbon PORTUGAL

Pontificia Universidad Católica de Chile CHILEUniversity of Sao Paulo BRAZILJoint Research Centre European Commission

Synthetic database (in collab. with G. Pirovano, RSE)

• Generated for PM2.5 using CAMx PSAT using the Po Valley dataset and

extracted in selected points

• The receptor corresponding to the city of Milan was used for the

intercomparison

• Emission profiles in literature were used to obtain the concentration of

trace elements

• Total sulphate, nitrate and ammonium calclulated to make them

comparable to those measured in monitoring sites

• Dust resuspension profile modulated using the wind speed measured in

the area

• Noise has been introduced using a random sampling from a log-normal

distribution

Intercomparison second step: preliminary results

Z scores: conclusive remarks (1)

• The previous evaluations were made using as reference a value that represents the average of all participants. In this test we use the same algorithm but the reference value represent the actual contribution added when constructing the database.

• In the test with an external reference value participants/solutions with median values in the optimum area (<1) are 18 out 26.

• Factor/sources in the acceptability area (<=2) are 85%, there is a 6% in the warning area and a 9% in the action area.

Z scores: conclusive remarks (2)

• In this test, FA-MLRA median is in the warning area. And the the median of CMB robotic solutions is close to the warning area.

• The advantage of this test is that is not necessary to remove solutions to avoid distortions. In addition, it is possible to assess possible bias of the solution that are not evident when using the average.

European Common Technical Protocol for Receptor Modelling

• C.A. Belis1, F. Karagulian1, B. R. Larsen2, F. Amato3,4, O. Favez5, I. El Haddad6, R.M. Harrison7, A.S.H. Prevot6, U. Quass8, R. Vecchi9, M. Viana3, P. Paatero10, P.K. Hopke11

1European Commission, Institute for Environment and Sustainability - Joint Research Centre, Ispra, 21027, Italy2European Commission, Institute for Health and Consumer Protection - Joint Research Centre, Ispra 21027, Italy3Institute of Environmental Assessment and Water Research (IDAEA), Barcelona, 08034, Spain4TNO Buil Environment and Geosciences, Dept. Air Quality and Climate, TA 3508,Utrecht, The Netherlands5Institut National de l’Environnement Industriel et des Risques (INERIS), Verneuil-en-Halatte, 60550, France6Laboratory of Atmospheric Chemistry (LAC), Paul Scherrer Institut, Villigen, 5234, Switzerland 7National Centre for Atmospheric Science, University of Birmingham, Birmingham, B15 2TT, UK8Air Quality and Sustainable nanotechnology (IUTA), D- Duisburg, 47229, Germany9Department of Physics, Università degli Studi di Milano and INFN, Milan, 20133, Italy10University of Helsinki, Finland 11Center for Air Resources Engineering and Science, Clarkson University, Potsdam, NY 13699-5708, USA

Common Receptor Model Technical Protocol (RMTP) Driving elements

• The objective is to promote the best available operating procedures and to harmonize their application across Europe.

• Target : -practitioners involved in the model execution and in the interpretation of results- policy makers interested in the output of RMs for the design of abatement measures, -air quality experts and scientists non familiar with this methodology

• Different levels of complexity according to the reader skills• It is organized following a logical sequence• Contains tutorials, technical recommendations and check lists• It is not meant to report all the information but to provide relevant

references to the information sources• The technical protocol is a guide and cannot substitute training and

expertise

PART B: STANDARD RECEPTOR MODEL TECHNICAL PROTOCOL

Is the core of the document. Contains description of the steps required in the most traditional and widespread Receptor Modelling techniques with particular reference to CMB and Factor Analysis 

PART C: SOPHISTICATED AND ADVANCED MODELS

This section contains innovative and advanced methods most of which under continuous development. Also methods on trajectories that although have been available for long time their potentials have not been completely exploited

PART A: INTRODUCTION TO SOURCE APPORTIONMENT WITH RECEPTOR MODELS

Presents the work and provides the unskilled reader with basic elements on Source Apportionment and Receptor Modelling 

Common RMTP Outline

PART A: INTRODUCTION TO SOURCE APPORTIONMENT WITH RECEPTOR MODELS

1. Glossary

2. Structure of the Document

3. Identification of pollution sources

4. What are RM?

5. When to use RM?

6. Harmonization of RM

Common RMTP Outline

PART B: STANDARD RECEPTOR MODEL TECHNICAL PROTOCOL 

1. PRELIMINARY EVALUATION OF THE STUDY AREA

2. DEFINING A METHODOLOGICAL FRAMEWORK

3. EXPERIMENTAL DESIGN

4. DATA COLLECTION / FIELD WORK/ CHEMICAL ANALYSES

5. KNOWING YOUR DATASET: BASIC STATISTICS

6. PRELIMINARY DATA QUALITY CHECK

7. INPUT DATA UNCERTAINTY CALCULATION

8. CHEMICAL MASS BALANCE MODELS

9. FACTOR ANALYSIS I: SELECTION OF THE NUMBER OF FACTORS

10. FACTOR ANALYSIS II: EVALUATION OF SCE AND MODEL

PERFORMANCE INDICATORS

11. FACTOR ANALYSIS III: CRITERIA FOR FACTOR LABELLING

12. OTHER MODEL PERFORMANCE TESTS

13. REPORTING RESULTS

DATA PRE-TREATMENT

FIELD AND LAB WORK

PRELIMINARY ACTIVITIES

COMPLEMENTARY TESTS AND REPORTING

RECOMMENDATIONSSPECIFIC FORFACTOR ANALYSIS

RECOMMENDATIONS SPECIFIC FOR CMB

more details

PART C: ADVANCED MODELS

Combination of trajectories and wind direction analysis with receptor models makes it possible to evaluate the geographic provenience of sources. These techniques are also useful for RM output validation.

1. WIND AND TRAJECTORY

ANALYSIS IN SOURCE

APPORTIONMENT

These models represent the new frontier of RM since make it possible to combine different type of data and make advanced data treatment

2. CONSTRAINED

AND EXPANDED

MODELS IN FACTOR

ANALYSIS

Thanks to the application of (PMF) to find factors this methodology has many analogies with the traditional RMs. The experience gained by the community working with these tools may be useful for RM experts and viceversa.

3. THE USE OF PMF

IN AMS DATA

PROCESSING

This is a promising technique that opens the opportunity for mutual validation with traditional RMs.

4. THE AETHALOMETER

MODEL

5. ISOTOPIC RATIOS COMBINED WITH

MACROTRACER ENRICHMENT FACTORS

Reporting results

• The results are described according to the steps proposed in sections B1- B12.

• Expert decisions concerning the identification of the number of sources and factor assignment are described providing evidence of the objective information that support them.

• The quantitative uncertainty of the output should be estimated and reported using the embedded tests included in the most robust methodologies.

• In addition, estimation of overall uncertainty and validation should be achieved using Monte Carlo permutations approach, by comparing outputs from independent models on the same data set, or applying other validated techniques.

Conclusions

• The use of Receptor Models in Europe is widespread and there is a steady increasing trend.

• In addition, RMs are useful for the validation of other techniques• The Common RMTP aims at answering:

the need to inform and guide potential users (e.g. policy makers) the practitioners demand for common rules andthe request from final users of harmonization and quality assurance of SA output.

• The Common RMTP is complementary with the RMIE which provides information on the performance and uncertainty of these techniques.

• The Common RMTP is a support to practitioner that apply RM but cannot substitute experience and training

• The Common RMTP is a living document that needs to be updated according to the development of the methodologies.

Thank you for your attention