pomi kick-off meeting ispra, 7/3/2008 ninfa: air quality forecast over the po valley basin. marco...
TRANSCRIPT
POMI kick-off meeting
Ispra, 7/3/2008
NINFA:
Air quality forecast over the Po Valley Basin
.
Marco Deserti, Enrico Minguzzi, Michele Stortini, Giovanni Bonafè
Regione Emilia-Romagna
ARPA-SIM, Area Meteorologia Ambientale
Contents
• The NINFA modelling system• 1 year hindcast: apr 2003 – mar 2004
(model verification and scenarios)• Model intercomparison• 4 year hindcast: 2003-2006 (interannual
variability)
NINFA modelling system (1) Northern Italian Network to Forecast photochemical and Aerosol pollution
Orography height (m)
CTM: Chimere (dust & sea salt included)
Meteorological input : COSMO- IT (7 km horizontal resolution, under test 2.8 km)
NINFA BPA (operational):
•10 km horizontal resolution, 8 vertical levels up to 500 hPa (next 5 km)
•Emissions: adapted from Corinair 2000 Italy + EMEP
• Boundary conditions: Prev’air (0.5°*0.5°)
NINFA ER (not operational):
•5 km horizontal resolution,
•Emissions: from ER 2003 (next INEMAR)
• Boundary conditions: NINFA BPA
NINFA modelling system (2)
The Chimere CTM has been adapted to Northern Italy:
• interface with COSMO meteorological fields
• modification of MH and Kz evaluation
• more urban corrections to meteorological input
• evaluation of plume rise for point sources NOx gridded emissions, year 2000 annual total
The NINFA system is used for:
- operational air-quality forecasts and hindcast (started in October 2005, available at www.arpa.emr.it/sim)
- Long-term simulations for air-quality assessment and scenario evaluation (4-year hindcast simulation (apr 2003 – mar 2007), Meteorological input from COSMO-IT re-analysis)
Ongoing: Upgrade with the new Chimere version, from 10 to 5 km horizontal resolution
The Meteorological model
Re-analysis mode• Forecast runs have not enough parameters => re-
analysis• 12 hours run chain • same rotated grid of the forecast model, 7 km grid
pace, 35 vertical levels• BC: ECMWF analysis (every 6 hours) • IC first level: ECMWF analysis (to avoid deviation)• IC upper levels: previous LAMI run• Hourly nudging (Schraff and Buchold, 1999)
towards the measured Synop data during the model run in order to find the sweet spot between coherence and realism
COSMO-IT (formerly Lokal Modell - LAMI)• Multi-scale non-hydrostatic meteorological model (Steppler
et al., 2003)• Clouds and precipitation micro-physics• Convection, radiation, turbulence, interaction between
Earth surface, soil and atmosphere
see (http://cosmo-model.cscs.ch/public/various/operational/arpa/operationalAppsARPA.htm
Mixing height
Average mixing height during winter months, estimated by Chimere pre-processor. Default configuration (left) and setup adapted to Northern Italy (right)
PM10 annual average
winter: + 5 - 7 g/m3
Summer: + 4 - 5 g/m3
Marco Deserti:
Modifications to Hmix: disabled enhancement below clouds, modified nocturnal scheme (now Mahrt 1981, function of U* only), increased minimum value in urban cells, introduced a maximum value of 2500 m
-Plume rise scheme: taken from CAMx model (Turner 1986, modified )
Marco Deserti:
Modifications to Hmix: disabled enhancement below clouds, modified nocturnal scheme (now Mahrt 1981, function of U* only), increased minimum value in urban cells, introduced a maximum value of 2500 m
-Plume rise scheme: taken from CAMx model (Turner 1986, modified )
Modifications to Hmix: disabled enhancement below clouds, modified nocturnal scheme (now Mahrt 1981, function of U* only), increased minimum value in urban cells, introduced a maximum value of 2500 m
NINFA Northern Italian Network to Forecast photochemical and Aerosol pollution
• Run every day on a Linux work station. Start at 4:00 GMT, output available at 09:00 GMT.
• NINFA is based on the regional version of photochemical model CHIMERE developed at Ecole Polytechnique, Paris.
• Boundary conditions by Prev'air data (www.prevair.org).
• Emission input data from the Italian National Inventory (year 2000) adapted for the species required by the MELCHIOR photochemical mechanism.
• point source emissions: a plume-rise module has been added to CHIMERE pre-processor.
• Land use: detailed Italian Corine2000 and European GLC2000.
• A suitable interface was constructed, to build CHIMERE meteorological input files starting form LAMI output.
• fields from COSMO assimilation cycle (LAMA) are used for NINFA long-term analysis.
METEO: COSMO IT/LAMA
CHIMERE
EMISSIONS:CTN_ACE
BOUNDARY CONDITIONS:
(Prev'air)
OUTPUT:O3, NO2, SO2, PM10
LANDUSE: CORINE2000+GLC2000
The modelling suite
Prev’air (Chimere-continental 0.5°*0.5°)
Urban model (ADMS Urban)
NINFA: Northern Italian Network to Forecast photochemical and Aerosol pollution
NINFA BPA 10 km ris.
NINFA ER 5 km ris
Multiscale approach
NINFA (ER-Chimere-regional-Po valley domain)
Prev’air (Chimere-continental-Europe-domain)
CORINAIR 2000 (COVN ton/anno)Input meteo COSMO-IT
Boundary conditions from Prev’Air
The model domain has an extension of 640 km x 410 km, 10 km horizontal resolution, with eight vertical levels up to a height of 5000 m. This relatively coarse resolution allows the use of homogeneous emission inventories and meteorological data on the whole domain, and helps keeping computer times reasonably short.
Numerical Air Quality forecast for northern Italy
SUMMARIZING……..
• ARPA – SIM provide daily numerical air quality forecast over the Po valley basin by the NINFA integrated modelling system;
• NINFA is a main tool to prepare the subjective AQ forecast over the Emilia-Romagna Region;
• NINFA is also applied for long term runs (hincast by high resolution meteorological analysis, produced by the COSMO model assimilation cycle)
• hindcast results are stored and can be distributed (available April 2003 - Mar 2006),
• NINFA outputs provide boundary conditions for the high res. runs over the Emilia-Romagna (NINFA ER and Urban models).
NINFA and POMI
Disclaimer:• At present POMI is not recognized by ER as a joint AQ
assessment exercise.
Which could be the contribution from ER ?• Provide NINFA hindcast results already available (10
km res.)• Run NINFA with POMI emissions and COSMO-IT
meteo data (5 km res. possible);• Provide observations: AQ and meteo data (already
available by DEXTER)Topics to be better defined:• Goals of the exercise ? (model comparison/validation or model
ensemble ?) Which is the added value (after xx-Delta & CTN)?• How the results will be evaluated and reported?• For which purposes?• Which data (input output) will be available from POMI?
Some results
1-year hindcast simulation (apr 2003 – mar 2004), Meteorological input from COSMO (LAMA) re-analysis)
Model validation
Fonte: APAT- CTN-ACE 2004
(*) the accuracy for modelling is defined as the maximum deviation of the measured and calculated concentration levels, over the period considered by the limit value, without taking into account the timing of events.
Data-quality objectivesData-quality objectives
Pollutant Av. time
Data-quality objectives for
Modelling(*)
Data-quality objectives for
continuous measurement
Italian law EC Directive
SO2, NO,
NO2
1 h1 d1 y
50 – 60 %50 %30 %
15 %
DM 2 aprile 2002, N. 60
1999/30/EC
PM, lead 1 y 50 % 25 %
CO 8 h 50 % 15 %2000/69/EC
Benzene 1 y 50 % 25 %
O3, NO, NO2 1 h day8 h max
50 %50 %
15 % To be received 2002/3/EC
Data set:Data set:
51 stations:
•8 rural background
•24 urban background
•11 urban traffic
•6 suburban background
•1 urban industrial
•1 suburban industrial
Model validation: RESULTS
configurazione del modello BPA 10km BPA 10km ER 5km
stazioni di controllo inquinante indicatore
criterio di qualità della modellazione (DM 60 2002 e Dlgs 183 2004)
BPA ER ER PM10 media annuale stazioni con errore <50% 81% 88% 75% NO2 media annuale stazioni con errore <30% 79% 78% 78%
ozono (semestre estivo)
massimo giornaliero della media su 8 ore
dati con errore <50% 93% 92% 93%
Model verification
• daytime ozone concentrations (1-hour and maximum daily 8-hour mean) agree very well with the observed ones, with correlation coefficients higher than 0.7 and low bias
• PM10 annual mean levels are underestimated (the bias is approximately -20 μg/m3), although correlation coefficients for the daily mean are around 0.6
PM10, year 2003-2004
0,00
10,00
20,00
30,00
40,00
50,00
RB (N=2) SB (N=6) UB (N=8) UT (N=4)
mg/
m3
sim oss
O3, summer2003
0
20
40
60
80
100
RB (N=3) SB (N=8) UB (N=16) UT (N02)
g/m
3
sim oss
Models intercomparison: O3 summer period
Source: CTN-ACE report 2007
Models intercomparison: PM10 winter period
Source: CTN-ACE report 2007
O z o n o - G io rn o tip ico d a l 01 /06 /1995 a l 31 /08 /1995
0
20
40
60
80
100
120
140
160
1 3 5 7 9 11 13 15 17 19 21 23
O R E
g/m 3
città
m ontagna
co llina
rura lip ianura
F igura 1: confronto tra i giorni tip ici estivi d i quattro d iversi siti d i m isura durante il periodoestivo 1995.
NINFAmodel
April – Sept 2003
The daily cycle is well reproduced by NINFA:
Plane: high peak values during the day, minimum during the night,
Mountain: little diurnal cycle…
Observed: MOTAP
Model verification: Ozone mean day in the plane, in the hills and in the mountain
Model verification: PM10 speciation
Bologna, annual mean*Warning: 2003 vs 2003-2004 !
* Data from CNR-ISAC (S.Fuzzi, C. Facchini)
Composizione PM10 a Bologna
NINFA Estate %
Primario35%
Nitrati5%
Ammonio8%
Solfati17%
Organico16%
Risosp19%
Sale0%
PolveRE Estate
Insolubile60%
Nitrati9%
Ammonio1%
Solfati18%
Cloruri0%
altro12%
NINFA Inverno
Primario36%
Nitrati35%
Ammonio13%
Solfati10%
Organico3%
Risosp3%
Sale0%
PolveRE Inverno
Insolubile41%
Nitrati32%
Ammonio7%
Solfati12%
Cloruri3%
altro5%
PM10 speciation and PM size distribution in Bologna
There is a lack of experimental data, a very rough comparison indicate that:
• organic seems to be strongly underestimated
• Inorganic is underestimated
• Dust agreement• Salt: sea salt can be
neglect in Bologna, other sources..?
• Similar results for continental (Prev’air 50 km) and regional (NINFA 10 km) simulations
• There is a general, although rough, agreement between observed and simulated size distribution
PM size distribution normalized annual mean
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0,35
0,40
0,45
0,50
0.01-0.6 0.6-2.5 2.5-10
D (micron)
mic
rog
/m3
NINFA POLVERE
PM speciation, Bologna
0,00
2,00
4,00
6,00
8,00
10,00
NO3 -hno3
NH4 -nh3
nssSO4- hso4
OM -soa
min.dust
sea salt unacc. -ppm
g/
m3
Putard 2003 Long4 Prev'air
COSMO IT: Some problems
Wind velocity 10 m, BIAS frequency distribution, Thermal inversion strength (00GMT), frequency distribution, S.Pietro Capofiume station
Strong nocturnal inversions are underestimated
Wind calm are underestimated
385
258
314
238
132 147
0
50
100
150
200250
300
350
400
450
COV COV
Lombardia Emilia-Romagna
kTo
n/a
rea/
ann
o CityDelta97
CityDelta CLE2000
INEMAR 2001
CTN 2000
ER 2000
232208
225
183
104
132
0
50
100
150
200
250
NOx NOx
Lombardia Emilia-Romagna
kT
on
/are
a/a
nn
o CityDelta97
CityDelta CLE2000
INEMAR 2001
CTN 2000
ER 2000
22
36
24 23
14 12
05
10152025303540
PM10 PM10
Lombardia Emilia-Romagna
kTo
n/a
rea/
ann
o CityDelta97
CityDelta CLE2000
INEMAR 2001
CTN 2000
ER 2000
EMISSIONS:
Annual total from different data sources over Lombardia and
Emilia Romagna regions
CityDelta: http://aqm.jrc.it/citydelta/
CTN 2000: http://www.sinanet.apat.it
LombardiaNOx COV PM10(kt/anno) (kt/anno) (kt/anno)
CityDelta97 232 385 22CTN 2000 183 238 23
difference % 27 62 -4
CityDelta CLE2000 208 258 36CTN 2000 183 238 23
difference % 14 8 57
INEMAR 2001 225 314 24CTN 2000 183 238 23
difference % 23 32 4
Emilia RomagnaNOx COV PM10(kt/anno) (kt/anno) (kt/anno)
ER 2000 132 147 12CTN 2000 104 132 14
difference % 27 11 -14
ER 2003 129 142 11CTN 2000 104 132 14
difference % 24 8 -21
Air quality assessment in Northern Italy
• NINFA has been run over 1 year period in the hindcast mode to simulate ozone and PM10 concentration.
• The hindcast run is helpful to estimate the size of the polluted area and to analyze the spatial patterns of the atmospheric pollution in Northern Italy
• The spatial structure of the simulated fields reproduces the mountain-plain concentration gradients of pollutants.
• spatial patterns are linked to wind regimes, characterized by frequently stagnation of air masses in the Po Valley and to the emissions distribution,
• Ozone: large amount of exceedances of the target value for the protection of human health (120 μg/m3 maximum daily 8-hour mean) are in the sub alpine region and in the plane. Most exceedances (up to 120 per year) are located downwind of the main urban agglomerates (Milano and Torino).
• PM10: annual average reaches its highest value in the plain area, extending from the west sub alpine region to the North-East Adriatic coast. The highest values are located in and around the main urban agglomerates (Milano and Torino).
4-year hindcast simulation (apr 2003 – mar 2007)
Objectives:
• Study the interannual variability in air quality due to meteorological conditions
• Remove the meteorological variability from observed concentrations to see if there is a real trend in emissions
• Investigate the uncertainty in emission reduction scenarios introduced by meteorological variability
Focus on fulfilment of EU legislation requirements for air quality (maximum 8h average for O3, daily average for PM10)
Background:
Most studies on air quality assessment and emission reduction scenarios (eg. City-Delta), are based on annual CTM simulations.
The particular year to be simulated is normally chosen “a priori”, mostly depending on data availability
7th EMS Annual Meeting7th EMS Annual Meeting
Results: OResults: O33
• Summer 2003 is exceptional (especially the number of exceedances)• The model reproduces very well the differences between years (only a small overestimation of day-time average)• Inter-annual variability is about 20% for average and 40% for exeedances • Model bias is constant in the “real world” there is no significant change in emissions
O3, summer, day hours, all stations: average concentrations (left) and number of days with 8h average > 60ppb (right) in different years
Results: PM10Results: PM10
• Annual average concentrations are almost constant (summer compensate winter); inter-annual variability is less than 15%• Observed variability in seasonal average can be explained by meteorology alone (no appreciable effect of changes in emissions)• Model underestimation is rather homogeneous in time
PM10, average concentrations in different years: winter months (left) and summer months (right)