1. CITY-DELTA Objectives, Methodology, and Results
C. Cuvelier, P. Thunis, L. Rouïl
on behalf of the steering group

2. Contents Background and Objectives Methodology
Interpretation of the results
Emission Inventories
Model Validation
Deltas Interpretation
Conclusions
Functional Relationships

3. Contents Background and Objectives Methodology
Interpretation of the results
Emission Inventories
Model Validation
Deltas Interpretation
Conclusions
Functional Relationships

4. Driving forces of health impacts
WHO systematic review of health impacts from air pollution (2003/2004):
Largest damage from long-term exposure to PM2.5
- Not yet possible to distinguish potency of individual PM components
- No threshold can be identified
- Thus larger health benefits from large-scale reductions of low concentrations than from peak concentrations at hot spots
New evidence for mortality effects from ozone
- From time series studies
- No firm evidence for no-effect level, but larger uncertainties for effects at low concentrations
- Thus also low ozone days are relevant

5. Modelling questions for CAFE
What are the downscaling effects when a refined grid is used over cities instead of “regional” grid
How to include sub-grid effects into an Europe-wide health impact assessment for PM/ozone?
How to balance Europe-wide vs. local emission controls to reduce health impacts in a cost-effective way?
NOT: How to model compliance with AQ limit values of AQ-Directives!

6. Contents Background and Objectives Methodology
Interpretation of the results
Emission Inventories
Model Validation
Deltas Interpretation
Conclusions
Functional Relationships
Discussion

7. City Delta A model inter-comparison exercise for
urban/regional dispersion models to identify:
Systematic deltas between rural and urban AQ  Scale
How these deltas depend on emissions  Emissions
How these deltas vary across cities  Cities
How these deltas vary across models  Models
How these deltas vary for PM and O3  Pollutants

8. 7 European-scale and … … 11 urban-scale models Model # of levels Level
Domain
Resolution
CALGRID 11
10m
CITYDELTA
5-10
CAMx
CHIMERE local
6: surface-700hPa
SL (50m) 5
CHIMERE regional
Europe 50
EMEP Unified Model
20: surface-100hPa 1m
EMEP
EMEP-v.1
45m
EPISODE
6: m 2m 10
EUROS 4
25m
10-50
LOTOS local
3: m ML
LOTOS régional
MOCAGE
47: surface-5hPa
1st level (0-50m)
Paris, Milano
MUSCAT
22: m
1st level (0-33m)
MUSE
OFIS 2
REM3 local
4: m SL
REM3 regional
STEM-FCM
TRANSCHIM
50m

9. Cities: London Paris Prague Berlin Copenhagen Katowice Milan Marseille
Comparisons are conducted for a number of European cities with distinct differences in climatic conditions, geographical settings, and emission densities.
London
Paris
Prague
Berlin
Copenhagen
Katowice
Milan
Marseille

10. Input Data: Monitoring data: Meteorological data:
For the 8 cities delivered by the city-authorities:
O3, NO2, PM2.5, PM10
Meteorological data:
Provided to CityDelta by Meteo-France, or by the modelling
group themselves
Emission inventories:
High-resolution (1 km to 5 km) city-emission inventories
Low-resolution (50 km) EMEP-TNO emission inventory
Boundary conditions:
Provided by EMEP, or by the modelling group themselves

11. Emission Scenarios NOx 0 --- 1999 1 --- 2010 CLE: Current Legislation
CLE: Current Legislation
NOx MFR: Maximum Feasible Reduction
NOx (CLE+MFR)/2
VOC MFR
NOX and VOC MFR
PMcoarse MFR
PM2.5 MFR
NOx
CLE-1999
MFR-1999
Prague
-34%
-62%
Milan
-36%
-53%
Paris
-42%
-65%
Berlin
-38%
-50%

12. OUTPUT Hourly values for O3 (and NO2) for 6 months (Summer)
GAS PM
Hourly values for O3 (and NO2) for 6 months (Summer)
Daily values for PM2.5 and PM10 for 12 months

13. The JRC tool Monitoring data Validation with Obs. Scenario Deltas
2D Avg. Fields
Scenario Deltas
Emissions

14. Contents Background and Objectives Methodology
Interpretation of the results
Emission Inventories
Model Validation
Deltas Interpretation
Conclusions
Functional Relationships

15. Local vs EMEP Regional emission inventories
MILAN
NOx, CO, SOx estimates seems quite robust (uncertainty 20-30%)
PM estimates show larger than 50% differences.
CITY DELTA has contributed to a considerable revision of the regional emission data
and resulted in the development of a new methodology for distributing emission data that secures consistency among pollutants and a traceable improvement in the emission distribution by sector.

16. Taylor plots: Berlin city centre
CRMSE
Corr.
K.E.Taylor, 2001, JGR, 106,

17. The ENSEMBLE model OZONE PM10 Day1 Day2 Day3 Day4 ExcD60 MeanT

18. City variability
O3 MeanT (CD)
PM10 Winter (CD)

19. Summary of model validation
Concentration levels
Correlation
Deviation from mean
Difference between coarse and fine scale models
Range O3
+/- 20 %
No large differences, some additional titration with FS
NO2
0 to -80%
Strong underestimates disappear with FS
PM10
-20 to -50%
-3 to +20 μg/m3,

20. NOx Reduction
O3 SOMO (CD)
VOC Reduction

21. NOx Group Reduction
PM10 Winter (CD)
VOC Group Reduction

22. Key findings: Ozone
Model results are highly sensitive to the quality of the emission inventories.
Generally,
- models reproduce well the present situation,
- they agree on the ozone changes expected from CLE in 2010
Coarse and fine scale models - when averaged over the regional scale - agree over wide ranges of responses of ozone to emission controls.
Fine-scale models are able to show important sub-grid effects, which are not captured by regional scale models.
Models agree more on the response to VOC emission controls than on the effects of NOx cuts (titration effect modelling).
The use of the ENSEMBLE model response provides a robust tool for analyzing the impacts of further emission reductions.

23. Key findings: PM
Validation of PM is hampered by the lack of reliable (chemically speciated) observations
All models underestimate total PM mass, both at urban and large scale. This is due to limited understanding of sources and processes.
Models agree that a large part of PM found in urban background originates from the regional background.
Urban PM levels can – to a large extent - be related to the density of primary urban low-level emissions, regional background, and city characteristics (e.g., ventilation).

24. Contents Background and Objectives Methodology
Interpretation of the results
Emission Inventories
Model Validation
Deltas Interpretation
Conclusions
Functional Relationships

25. Functional relationships: Basic Approach
50km
City
50km
Delta Conc = PMconc in a given 5x5km Grid - Average PMconc over whole Domain
Delta Emis = ED in a given 5x5km Grid - Average ED over whole Domain
Correlate: Delta Concentration vs Delta Emission Density

26. ENSEMBLE Base Case, CLE and MFR “Conc-Emis” correlations
PARIS
Paris: Slope = 0.23
R2 = 0.82
MILAN
Milan: Slope = 1.64
R2 = 0.68

27. Slopes of individual cities against EMEP wind speed in city grid

28. Functional relationship for PM
DPMsub-grid = (EDsub-grid - EDEMEP) * (k1 - k2*Vwind)
DPMsub-grid .. Difference in PM concentration between sub-grid (urban/rural) area and EMEP grid average
EDx … Emission density for low sources (x=urban/rural/EMEP grid average)
Vwind … Annual mean wind speed in EMEP grid cell
k1, k2 … Parameters derived from the City-Delta ensemble model

29. Modelling concept for urban background PM2.5
Take grid-average (50*50 km) PM2.5 concentrations from the regional scale EMEP model
These include:
primary PM and
secondary inorganic aerosols
and do not include:
secondary organic aerosols
PM from natural sources
Add urban (sub-grid) increment, based on City-Delta functional relationships

30. PM implementation in RAINS
ΔPMsub-grid = (EDsub-grid – EDEMEP) * (k1 - k2*Vwind)
= (EDEMEP * (PDsub-grid / PDEMEP ) – EDEMEP) * (k1-k2*Vwind) = EDEMEP * (PDsub-grid / PDEMEP – 1)* (k1 - k2*Vwind)
Necessary input data:
Emission densities of low level sources in an EMEP cell (for an emission control scenario)
Population densities in urban and rural areas for an EMEP cell
Annual mean wind speed in an EMEP grid cell

31. Population density ratios
Input data
Wind speed
Emission densities
Population density ratios
Urban increments

32. Anthropogenic contribution to PM2. 5 Grid average vs
Anthropogenic contribution to PM2.5 Grid average vs. urban increments, 2000 [µg/m3]

33. Validation against observations Urban background PM2.5 [μg/m3]

34. Discussion
Urgent need for validation with monitoring data, hampered by lack of PM2.5 twin sites.
Presently, grid average wind speed used. No consideration of topography. City-specific wind speeds should be improved.
Which emission/population density is representative for a city (how to draw city borders)? This determines directly the size of the urban increment.
For which city size are the FR applicable? (introduce lower limit)

35. Conclusions (1)
A first approach for addressing urban air quality for Europe-wide health impact assessment has been developed and implemented – based on observations and City-Delta results
Major determinants:
Regional background
Urban emission densities of low level sources (traffic, heating)
Meteorological and topographical information (wind speed)
Geographical information (size of city, population)

36. Conclusions (2)
First results are promising, further refinement is necessary
More PM2.5 monitoring data is necessary for validation
Uncertainty and sensitivity analyses not yet performed
Only for health impact assessment, not yet suitable for compliance with AQ limit values