FAIRMODE
The combined use of models and monitoring
for applications related to the European air
quality Directive: SG1-WG2 FAIRMODE
Bruce Denby
Wolfgang Spangl
FAIRMODE 3rd Plenary, Kjeller Norway
September2010
Content
•
•
•
•
•
Aims of SG1-WG2
Overview of methods
Institute and activities list
Registered FAIRMODE list
Agenda and plans for today
Aims of SG1
• To promote ‘good practice’ for combining models
and monitoring (Directive related)
• To provide a forum for modellers and users
interested in applying these methodologies
• To develop and apply quality assurance practices
when combining models and monitoring
• To provide guidance on station representativeness
and station selection
Some concepts
• ’Combination’ used as a general term
• Data integration
– Refers to any ‘bringing together’ of relevant and useful information
for AQ modelling in one system (e.g. emissions/ meteorology/
satellite/ landuse/ population/ etc.)
• Data fusion
– The combination of separate data sources to form a new and optimal
dataset (e.g. models/monitoring/satellite/land use/etc.). Statistically
optimal but does not necessarily preserve the physical characteristics
• Data assimilation
– The active, during model integration, assimilation of observational
data (e.g. monitoring/satellite). Physical laws are obeyed
Increasing statistical expertise
Expertise required for methods
Bayesian
heirarchical
approaches
Kriging
methods
Monte Carlo
Markov Chain
Optimal
interpoaltion
4D var
Ensemble
Kalman filter
Data assimilation
GIS based
methods
Regression
IDW
Data fusion
Modelling
Increasing model expertise
Users and developers (DA)
Person
Institute/project
Contact
Model
Method
Hendrik Elbern
he@eurad.Uni-Koeln.DE
EURAD-IM
3-4D var
Martijn Schaap
RIU/MACC/PASA
DOBLE
TNO/MACC
martijn.schaap@tno.nl
LOTOS_EURO
S
Ensemble Kalman
filter
L. Menut
INERIS/MACC
menut@lmd.polytechniqu
e.fr
CHIMERE
Hilde Fagerli
Met.no/MACC
hilde.fagerli@met.no
EMEP
Optimal
interpolation ,
residual kriging and
EnKF (in
development)
3 – 4D var (in
development)
Valentin
Foltescu
Sébastien
Massart
Bruno Sportisse
SMHI/MACC
MATCH
CERFACS/MACC
Valentin.Foltescu@smhi.s
e
massart@cerfacs.fr
INRIA,CEREA
Bruno.Sportisse@inria.fr
Jeremy David
Silver
Jørgen Brandt
NERI
jds@dmu.dk
jbr@dmu.dk
MOCAGE/PAL
M
Polyphemus
DEHM
2 – 4D var (in
development)
3 -4D var
Application
(resolution)
European forecasts
(45 – 1 km)
European
assessments and
forecasting (25km)
European and
Urban scale
forecasts and
assessments (25
km)
European scale
forecasts and
assessment (25km)
European to Urban
scale (25 - ? km)
Global to European
3 -4D var, OI, EnKF
European
Optimal
interpolation. Future
EnKF
Europe
Users and developers (DF:1)
Person
Institute/project
Contact
Model
Method
John Stedman
AEAT
John.stedman@aeat.co.uk
ADMS
Bruce Denby
NILU/ETC-ACC
bde@nilu.no
EMEP, LOTOSEUROS
Jan Horálek
CHMI/ETC
horalek@chmi.cz
EMEP
Dennis
Sarigiannis
JRC Ispra
Dimosthenis.SARIGIAN
NIS@ec.europa.eu
Marta Garcia
Vivanco
Palomino
Marquez
Inmaculada
Fernando Martín
CIEMAT
m.garcia@ciemat.es
inma.palomino@ciemat.e
s
fernando.martin@ciemat.
es
CTDM+ (model
not important,
platform more
relevant)
ICAROS NET
MELPUFF
CHIMERE
Statistical
interpolation,
residual kriging
Statistical
interpolation,
residual kriging
Statistical
interpolation,
residual kriging
Data fusion
(unknown
methodology)
Anisotropic inverse
distance weighting
Regression and
residual kriging.
Application
(resolution)
UK wide
assessment of air
quality
European wide
assessments at 10
km
European wide
assessments at 10
km
Urban scale
Assessment Spain
Users and developers (DF:2)
Person
Institute/project
Contact
Model
Method
Clemens
Mensink
Stijn Janssen
VITO
stijn.janssen@vito.be
Clemens.mensink@vito.b
e
RIO and
BelEUROS
J.A. van
Jaarsveld
Florian Pfäfflin
(Goetz Wiegand
Volker
Diegmann )
RIVM
hans.van.jaarsveld@rivm.
nl
fpf@ivu-umwelt.de
OPS
Detrended kriging.
Land use regression
model used for
downscaling CTM
Kriging with
external drift
Optimal
interpolation
Arno Graff
Umwelt Bundes
Amt, UBA II
Dipartimento di
Ingegneria
dell'Informazione
Università degli
Studi di Brescia
arno.graff@uba.de
REMCALGRID
TCAM
Gabriele
Candiani,
Marialuisa Volta
IVU Umwelt
GmbH
lvolta@ing.unibs.it
gabriele.candiani@ing.un
ibs.it
FLADIS/
IMMISnet/
EURAD
Optimal
interpolation
Optimal
interpolation,
kriging
Application
(resolution)
Belgium (3km)
Nederland (5km)
Ruhr, Germany
(5km)
Germany
Northern Italy
Representativeness
Person
Institute/project
Contact
Wolfgang
Spangl
Umweltbundesamt
Wolfgang.spangl@umwel
tbundesamt.at
Sverre Solberg
NILU/EMEP
sso@nilu.no
Model
Method
Representativeness of
monitoring data
EMEP
Representativeness of
monitoring data
Application
(resolution)
Austria
EMEP monitoring
network
Registered for SG1
N.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
Name
BRUCE DENBY (leader)
WOLFGANG SPANGL
BERTRAND BESSAGNET
HENDRIK ELBERN
THILO ERBERTSEDER
ALBERTO GONZÁLEZ
MARKE HONGISTO
FERNANDO MARTIN
CAMILLO SILIBELLO
MARTIJN SCHAAP
KEITH VINCENT
GABRIELE ZANINI
HILDE FAGERLI
AMELA JERICEVIC
CLARA RIBEIRO
HANS VAN JAARSVELD
MIREIA UDINA SISTACH
ONDREJ VLCEK
SUZANNE CALABRETTA-JONGEN
CAMILLO SILIBELLO
MARIALUISA
THILO ERBERTSEDER
RAHELA ZABKAR
DAVID CARRUTHERS
JOSE M. BALDASANO
MARIA GONCALVES-AGEITOS
GUUS VELDERS
GUUS STEFESS
e-mail
bde@nilu.no
wolfgang.spangl@umweltbundesamt.at
abriele.bessagnet@ineris.fr
he@eurad.uni-koeln.de
thilo.erbertseder@dlr.de
agortiz@mma.es
marke.hongisto@fmi.fi
abriele.martin@ciemat.es
c.silibello@aria-net.it
martijn.schaap@tno.nl
keith.vincent@aeat.co.uk
abriele.zanini@enea.it
h.fagerli@met.no
jericevic@cirus.dhz.hr
clararibeiro@ua.pt
hans.vanjaarsveld@pbl.nl
mudina@am.ub.es
vlcek@chmi.cz
suzanne.jongen@vito.be
c.silibello@aria-net.it
lvolta@ing.unibs.it
thilo.erbertseder@dlr.de
rahela.zabkar@fmf.uni-lj.si
david.carruthers@cerc.co.uk
jose.baldasano@bsc.es
maria.goncalves@bsc.es
guus.velders@pbl.nl
guus.stefess@rivm.nl
Agenda today in SG1-WG2
• Status
• Short presentation on representativeness (Wolfgang Spangl)
• Discussion around activities for this and next year:
– Defining representativeness for validation and assimilation purposes
– Overview of institutes applying assimilation methods
– Review of existing methods implemented in Europe. (Template and
contributions)
• Quality assurance and validation methods when combining
monitoring and modelling (Input to SG4)
• Financing and securing contributions to the activities of SG1
• 2011 activities (workshop, special session AQ conference)
• Any other business
Aims of SG1-WG2 2010
• Update list of institutes carrying out DA and DF
• Develop an accessible review of these methods
(→ WG1 ‘Good practise and guidance’)
– Contributions from listed institutes
• Recommend methods for quality assurance
(→ SG4 ‘Bench marking’ → WG1 ‘guidance’)
– Discussion today
• Develop an understanding of representativeness
(→ SG4 ’Bench marking’ → WG1 ‘guidance’)
– Discussion today
For information and contributions contact
Bruce Denby
bde@nilu.no
and register interest on the website
http://fairmode.ew.eea.europa.eu/
Some concepts
• Geometrical methods
– Methods for interpolation or ‘combination’ that are based on
geometrical arguments. E.g. Inverse distance weighting, bilinear
interpolation, as an interpolation method. Simple combinations of
data, some GIS based methods.
• Non spatio-temporal statistical methods
– Covers methods such as regression and bias corrections that do not
take into account the spatial or temporal correlation of the data.
• Spatio-temporal statistical methods
– Covers a wide range of methods e.g. 2-4 D variational methods,
kriging methods, optimal interpolation. Based on Bayesian concepts.
Minimalisation of some specified error.
Examples: Regional scale
Comparison of Residual kriging and Ensemble Kalman Filter for
assessment of regional PM10 in Europe
Residual kriging
EnKF
Model (LOTOS-EUROS)
Denby B., M. Schaap, A. Segers, P. Builtjes and J. Horálek (2008). Comparison of two data
assimilation methods for assessing PM10 exceedances on the European scale. Atmos. Environ. 42,
7122-7134.
Examples: Regional scale
Comparison of Residual kriging and Ensemble Kalman Filter for
assessment of regional PM10 in Europe
Residual kriging
EnKF
Model (LOTOS-EUROS)
Denby B., M. Schaap, A. Segers, P. Builtjes and J. Horálek (2008). Comparison of two data
assimilation methods for assessing PM10 exceedances on the European scale. Atmos. Environ. 42,
7122-7134.
Examples: Regional scale
MACC ensemble forecast system
EPS Graph
http://www.gmes-atmosphere.eu/.
Examples: Regional scale
MACC ensemble forecast system
Model
Assimilation method
Implementation
CHIMERE
Innovative kriging, Ensemble
Kalman filter
Not implemented in operational forecasts
EMEP
Intermittent 3d-var
In development
EURAD
Intermittent 3d-var
Implemented in forecast, using ground based
observations and satellite derived NO2
LOTOSEUROS
Ensemble Kalman filter
Not implemented in operational forecasts
MATCH
Ensemble Kalman filter
In development
MOCAGE
3d-FGAT and incremental 4dVAR
Not implemented in operational forecasts
SILAM
Intermittent 4d-var
Not implemented in operational forecasts
http://www.gmes-atmosphere.eu/.
Examples: Local and urban
• Few examples of data fusion/assimilation on the
local and urban scale
– Spatial representativeness of monitoring sites is very
limited (10 – 1000 m)
– Often the number of sites is limited (compared to their
spatial representativeness)
– Monitoring contains little information for initialising
forecasts
• Application for assessment is possible
– E.g. regression, optimal interpolation
Representativeness
• Two types of representativeness:
– spatial and temporal (physical)
– similarity (categorisation)
• Knowledge of this is important for:
– validation of models
– data fusion/assimilation
Representativeness
• For modelling applications the representativeness of
monitoring data should be reflected in the
uncertainty of that data
– NB: Not just the measurement uncertainty
• This is reflected in the AQ Directive (Annex I)
“The fixed measurements that have to be selected for comparison with
modelling results shall be representative of the scale covered by the
model”
• Representativeness will be pollutant and indicator
dependent
Representativeness and the AQD
• For monitoring the AQ Directive states:
– For industrial areas concentrations should be representative of a
250 x 250 m area
– for traffic emissions the assessment should be representative for
a 100 m street segment
– Urban background concentrations should be representative of
several square kilometres
– For rural stations (ecosystem assessment) the area for which the
calculated concentrations are valid is 1000 km2 (30 x 30 km)
• These monitoring requirements also set
limits on model resolution
Defining spatial representativeness
• The degree of spatial variability within a specified
area
– e.g. within a 10 x 10 km region surrounding a station the variability
is  30%
– Useful for validation and for data assimilation
• The size of the area with a specified spatial
variability
– e.g. < 20% of spatial mean (EUROAIRNET) or < 10% of observed
concentration range in Europe (Spangl, 2007)
– Useful for determining the spatial representativeness of a site
Observed spatial variability
Coefficient of variation σc/c for annual indicators as a
function of area (diameter) for all stations (Airbase)
0.5
Variability of mean NO 2 (2006)
Variability of mean PM10 (2006)
Variability of SOMO35 (2006)
0.5
0.5
0.5
0.45
0.45
0.45
0.4
0.4
0.4
0.3
At 5km resolution
variability is 34%
0.2
0
20
40
60
80
100
120
140
0
diameter
PM10
0.35
0.3
SOMO35
0.35
0.3
24%
0.25
160
180
200
0.2
Lag distance (km)
σc/c
Coefficient of variation
NO2
0.35
0.25
0.2
Coefficient of variation
Coefficient of variation
47%
0.25
0
20
40
60
80
100
120
140
160
180
200
0.2
0
20
40
Lag distance (km)
200 km
A random sampling within a 5km grid in an
average European city will give this variability
60
80
100
120
Lag distance (km)
140
160
180
200