Vector Search -An Introduction in Oracle Database 23ai.pptx
Interconfronto Europeo Particolato Atmosferico
1. QUINTO CONVEGNO
NAZIONALE
SUL PARTICOLATO
ATMOSFERICO
Interconfronto europeo
per modelli a recettore:
risultati preliminari
(Perugia 16-18 May 2012)
F. KARAGULIAN, C.A. Belis, F. Amato, D.C.S. Beddows, V. Bernardoni, S. Carbone, D. Cesari, E.
Cuccia, D. Contini, O. Favez, I. El Haddad, R.M. Harrison, T. Kammermeier, M.Karl, F. Lucarelli,
S.Nava, J. K. Nøjgaard, M. Pandolfi, M.G. Perrone, J.E. Petit, A. Pietrodangelo, P. Prati, A.S.H.
Prevot, U. Quass, X. Querol, D. Saraga, J. Sciare, A. Sfetsos, G. Valli, R. Vecchi, M. Vestenius, J.J.
Schauer, J.R. Turner, P. Paatero, P.K. Hopke
Joint Research Centre – IES - ACU
2. Real world data base (DB) choice
Site Location:
St. Louis Supersite (USA)
Merged two DBs with inorganic
and organic data with different
time resolution
(every day vs. every 6th day)
DB contained 178 samples
spanning two years
Missing values and values BDL
were already treated in
the inorganic DB,
but not in the organic DB.
Some data treatment was
asked to the participants
•MISSOURI
•ILLINOIS
3. Intercomparison participants
ORGANIZATION
COUNTRY
IDAEA CSIC
SPAIN
Univ. Aahrus
DENMARK
University of Genoa
ITALY
Finnish Meteorological Institute
FINLAND
INERIS/LSCE
FRANCE
University of Birmingham
UNITED KINGDOM
Norwegian Institute for Air Research
(NILU)
NORWAY
Department of Physics University of
Florence
ITALY
University of Milan Bicocca
ITALY
C.N.R. Institute for Atmospheric Pollution
Research
ITALY
IUTA e.V.
GERMANY
NCSR Demokritos, Environmental Research
Laboratory
GREECE
Dept. of Physics - University of Milan
ITALY
Paul Scherrer Institut Laboratory of
Atmospheric Chemistry
SWITZERLAND
C.N.R - I.S.A.C.
ITALY
JOINT RESEARCH CENTRE
UE
MODEL
PMF3-EPA
PMF-2
CMB
APCS
COPREM
ME-2
PCA
TOTAL
SOLUTIONS
8
6
4
1
1
1
1
22
16 PARTICIPANTS
4. DB compostion:
mass concentrations of species
and uncertainties
OCT
OC1
OC2
OC3
OC4
OP
ECT
EC1m
EC2
EC3
SO4
NO3
NH4
Al
As
Ba
Ca
Co
Cr
Cu
Fe
Hg
K
Mn
Ni
P
Pb
Rb
Se
Si
Sr
Ti
V
Zn
Zr
INORGANIC DB
From June 2001 – May 2003
24h samples collected every day
Reference
Lee, J. H., P. K. Hopke, and J. R. Turner (2006),
Source identification of airborne PM2.5 at the St.
Louis-Midwest Supersite,J. Geophys. Res., 111,
D10S10,
indeno(cd)pyrene
benzo(ghi)perylene
benz(a)anthracene
benzo(a)pyrene
fluoranthene
pyrene
ORGANIC DB
From May 2001 – July 2003
24h samples collected every day
Reference
coronene
benzo(b,k)fluoranthene
benzo(e)pyrene
benzo(j)fluoranthene
dibenz[a,h]anthracene
levoglucosan
Jaeckels JM, Bae M.S., Schauer JJ
(2007) Positive matrix factorization
Analysis of molecular markers
measurements to quantify the sources
of organic aerosols. EST. 41-5763
Structure of errors:
inorganic ions: high uncertainty
Co, Cr, Hg, Ni, Rb, Ti, Va, Zr have many missing values
Ca, Fe, Zn, K uncertainties below 5%
there were differen MDLs, probably due to different
analytical batches
PAHs presented many BDL values.
5. Purpose of the work
Participants performed receptor modeling using one or more model approach
Complete DB with uncertainties was provided to each participant
- MDLs and analytical uncertainties were provided to allow estimation of uncertainties
(upon participant’s choice)
- Emission inventory and source profiles (SPECIATE) were provided for CMB’s users
According to participants’ results, 15 source types
were identified and compared
Biomass burning
Gasoline
Diesel
Brakes
Traffic
Dust
Sulphates
Nitrates
Zn Smelter
Cu metallurgy
Pb smelter
Steel processing
Industry & Combustion
Ship emissions
Secondary inorganic sources
6. • Preliminary Test:
• evaluate if the factors within each source
category are homogeneous
• Proficiency test:
• evaluate if the quantitative source contribution
estimations (SCE) fall within an established
quality objective
7. Summary preliminary test
Test to identify the correspondence of
participants factor profiles to each source category
FACTORS
Pearson raw data and log transformed
and weighted difference (WD)
FACTOR VS MEASURED SOURCES
Pearson raw data and log transformed
and weighted difference (WD)
If 4 out of 7 tests were nor meet
then
factor was considered dubious
CONTRIBUTIONS (TIME TRENDS)
Pearson raw data
Z’(SCE)
New methodology to compare
different chemical profiles
Participants’
performance
Model
paerformance
9. INTER-COMPARISON methodology I:
Weighted difference (WDij)
1. Weighted difference (WD) between two factors was calculated using
the following equation:
WDij = 1/n
n
∑
xia − x ja
a =1
2
s ia + s 2
ja
where xi and xj are the relative concentrations of the n species in the
profiles i and j, respectively, and si and sj are their uncertainties.
2. The range of acceptability for the
weighted difference was set between
0 and 2.
4.0
WDij
4.0
3.0
3.0
2.0
2.0
1.0
0.0
1.0
0.0
OK
NOT OK
10. Proficiency test (ISO 13528)
Defining an assigned (reference) value X (source contribution estimation SCE) and its
uncertainty uX as reference value to compare with participant’s run average xi.
z' (SCE) =
xi − X
(
u2 + σ p X
X
)2
Defining the standard deviation for proficiency assessment (σ p) as criterion to
‘
evaluate participants’ performance (ISO 13528) (σ p = 50%, total mass annual mean)
SCE of participant’s source profile are optimal if:
z ‘≤ 1
“OK”
considered coherent and satisfactory if:
1 < z ‘≤ 2
“acceptable”
results are considered questionable if:
2 < z ‘≤ 3
“Warning”
results are unsatisfactory if:
‘
z >3
“Action”
11. Preliminary screening
number number of
of factors participants
6
4
7
2
8
4
9
4
10
2
11
3
12
2
13
1
EPA PMF-3.0 PMF-2
8
6
PCA
ME-2
COPREM
CMB
APCS
1
1
1
4
1
average number of solution ranged from 7 and 11
PMF-3 = EPA PMF 3.0
No tri-linear PMF model!!!!
12. Modeled PM2.5 mass vs measured PM2.5 mass
PM total mass = 18000 ng/m3
high intercept and low slope
0.8 < R2 < 1
0.7 < R2 < 0.8
R2 < 0.7
High intercept and high slope
Low intercept and slope close to 1
13. Source Categories
Sources categories identified by the majority of the participants:
1. Biomass Burning (22),
2. Dust - Re-Suspended Soil (21),
3. Traffic (16),
4. Industry & Combustion (16)
5. Cu metallurgy (14)
6. Zn-smelter (11),
7.Sulphates (10)
8. Nitrates, Diesel (9)
9. Pb-smelter, Steel, Secondary (8)
10. Gasoline, Brakes, ships (<=6)
18. Conclusion I
1) The methodology used for the evaluation of the IE appears
effective to test the comparability between factors in terms of
both chemical composition and time trend.
2) The weighted difference is useful to provide an independent
estimation of the comparability between factors and makes it
possible to check if the uncertainties have been correctly
estimated.
3) There is a reasonable quantitative agreement between SCE. 90%
of the factors meet the acceptability criteria (OK or acceptable).
4) The participant bias in the SCEs appears to be consistent with the
50% maximum uncertainty acceptability criterion used in this
evaluation.
19. Conclusion II
6) Many of the factors are comparable with those reported by Lee &
Hopke in the original publication of results using only inorganic
species.
7) There is a considerable variability in the number of factors identified
by participants.
8) Some models were used by only one or two participants, therefore it
is not possible to draw conclusions about the performace of these
models.
9) The noise of the experimental data, the variety of methodological
approaches and the little knowledge about the sampling site shall be
taken into account when interpreting the intercomparison outcome.