Functional relationships for modelling urban pollution in RAINS/GAINS Results from the City-delta III project

1
Functional relationships for modelling urban
pollution in RAINS/GAINSResults from the
City-delta III project

• M. Amann, Z. Klimont, C. Heyes, W. Schöpp (IIASA)
• C. Cuvelier, P. Thunis (JRC), Artur Gzella
  (IIASA/JRC)
• L. White (LWA)

2
Acknowledgements

• A. Eliassen, L. Tarrason and P. Builtjes and W.
  Asman for their scientific input
• Concawe, BUWAL, DG-ENV, IIASA, JRC for their
  financial contributions
• The modelling teams of Chimere (L. Rouil, B.
  Bessagnet), CAMx (M. Bedoni, G. Pirovano), REM-3
  (R. Stern, A.Kerschbaumer) and TCHAM (C.
  Carnevale) for their hard work
• EEA-ETC for providing the Airbase AQ monitoring
  database
• LWA for processing the Airbase database
• JRC for providing data on population and wind
  speed

3
Contents

• Objectives
• General approach
• Hypothesis for the functional relationships
• Input data
• Derivation of parameters
• Extrapolation to all European cities
• Implementation in RAINS
• Validation against observations
• Initial projections for 2020
• Conclusions

4
Objectives of the City-delta project

• Quantify the influence of regional and local
  emissions on urban background pollution
• for Europe-wide health impact assessments,
• for PM2.5 and ozone,
• for all larger European cities,
• with available data,
• based on an ensemble response of state-of-the-art
  meso-scale dispersion models.
• Develop functional relations for use in the
  cost-effectiveness analysis in RAINS/GAINS

5
Definition

• Urban incrementIncremental (PM2.5)
  concentrations in a city originating from
  emissions of the same city (i.e., difference
  between PM2.5 concentrations in urban background
  air and at an upwind site)
• City-deltaCorrection of a PM concentration
  value computed by a 5050 km regional dispersion
  model to derive the background concentration
  within a city of that grid cell

6
Two hypotheses for urban PM2.5

• Only primary PM emissions from low-level sources
  increase PM2.5 concentrations within the city.
• The
• formation of secondary (inorganic and organic)
  aerosols, as well as
• the dispersion of primary PM2.5 emissions from
  high stacks
• occur on the regional scale and are reflected in
  the background computed by the regional-scale
  dispersion model.

7
The City-delta approach

1. Compile a data sample of model responses towards
   changes in urban emissions
2. Hypothesis of a functional form of a relationship
   
3. Regression of the coefficients
4. Extrapolation to 473 other cities with local
   explanatory variables Urban increments
5. Computation of the resulting City-deltas
6. Validation against observations

8
Data sample for the regression analysis

• 3 dispersion models (Chimere, CAMx, REM)
• 7 cities with different characteristics (Berlin,
  Krakow, Lisbon, London, Milano, Paris, Prague)
• For emissions in 2020 with and without urban
  emissions from low level sources
• For full year meteorology of 2004
• Output Computed decreases in urban background
  PM2.5 concentrations after switching off urban
  low-level emissions

9
Comparison of observed and computed PM2.5
concentrations
10
Summary of model responsesSwitching off urban
low-level emissions in 2020(when PM2.5 emissions
will be 40-60 lower than in 2000)
11
Hypothesis for an urban increment of PM2.5

• For neutral atmospheric stability General
  formula for concentration increment (?c) of a
  non-reactive pollutant from low-level sources
  (e.g., Seinfeld and Pandis, 1998)D
  diameter of city, U wind speed, ?Q emission
  rate
• For low-wind speed conditions
• Summer Formation of secondary organic aerosols
  is not included in models cannot be
  represented in FR.
• Winter Models show high contribution of low-wind
  speed days to annual mean PM2.5 concentrations.

12
Contribution of low-wind speed winter days to the
annual urban increment
13
Regression formula

• ?c concentration increment computed with the
  3 models
• a. ß regression coefficients
• D city diameter
• U wind speed
• ?q change in emission fluxes
• d number of winter days with low wind speed

14
Performance of functional relationships
15
Input data City shapes and diameters
IIASA city population location (center
points) data, gt 100,000 inhabitants based on
www.citypopulation.de and ArcEurope Base Map
JRC ISPRA population density data, 1 x 1 km.
16
City shapes and diameters

• 473 cities in Europe with more than 100.000
  inhabitants
• Including 60 of European population

17
Meteorological input data for FR
Annual wind speed
MARS meteorological Database (JRC). Based on 2000
Weather stations (Lambert Azimuthal 50x50 km2
grid).
18
Frequency of low wind speed days (lt1.5 m/s)Data
for 2004
19
Topographic factor (D/U)1/2 (proportional to
µg/m3 concentration increment per ton PM2.5
emissions under neutral atmospheric conditions)
20
Emission data

• Urban emissions have been extracted from the
  gridded EMEP inventory
• Gridded data for PM2.5 have been supplied to EMEP
  by AT, DK, ES, FI, FR, LT, UK (PM10)
• These national inventories are not based on urban
  inventories
• For all other countries EMEP has assumed domestic
  and transport PM2.5 emissions distributed
  proportional to population densities
• This implies wood burning also in cities
• Wood burning in cities banned in UK, FR, IT as of
  2005 (?)

21
Assumptions about urban emissions

• Low level sources
• SNAP 2 (non-industrial combustion, domestic)
• 50 of SNAP 3 (industrial combustion)
• 50 of SNAP 4 (industrial non-combustion
  processes)
• SNAP 7 (traffic)
• SNAP 9 (waste)
• Composition of low-level emissions in the 473
  CD-cities
• Wood burning 34
• Other domestic sources 9
• Industrial combustion 8
• Industrial processes 11
• Traffic 35
• Waste 3

22
Emission densities of urban low level sources
2004based on the gridded EMEP 2004 inventory
23
Per-capita PM2.5 emission from urban low-level
sources 2004 as contained in the gridded EMEP
2004 inventory
24
Computed urban incrementsfor 2004, based on
gridded EMEP emissions
Mean concentration change (µg/m3) computed over Mean concentration change (µg/m3) computed over Mean concentration change (µg/m3) computed over
1515 km 1010 km 55 km
Vienna 4.5 7.0 9.4
Berlin 1.6 2.5 3.4
Krakow 4.4 6.6 8.4
Lisbon 5.8 9.1 12.4
London 2.4 3.8 5.2
Milano 10.8 15.3 17.7
Paris 5.3 8.3 11.2
Prague 4.0 6.3 8.5
25
Computed urban incrementsfor 2000, based on
gridded EMEP emissions, for 1010 km area
26
Computation of the City-delta

• To avoid double-counting of urban emissions, the
  urban emission contained in the regional-scale
  model calculations must be subtracted
• The resulting City-delta CDC can then be added to
  the EMEP regional scale results PMEMEP to attain
  total PM concentrations in urban areas PMC

27
Validation for 2004 (AT, BE, BG, CZ, CK, EE,
FR)GAINS (Background EMEP City-delta)
estimates for PM2.5
28
Projection for 2020 (AT, BE, BG, CZ, CK, EE, FR
)GAINS (Background EMEP City-delta)
estimates for PM2.5
29
Validation for 2004 (DE, GR, HU, IE)GAINS
(Background EMEP City-delta) estimates for
PM2.5
30
Projection for 2020 (DE, GR, HU, IE)GAINS
(Background EMEP City-delta) estimates for
PM2.5
31
Validation for 2004 (IT, LV, LT, NL, PL,
PT)GAINS (Background EMEP City-delta)
estimates for PM2.5
32
Projection for 2020 (IT, LV, LT, NL, PL,
PT)GAINS (Background EMEP City-delta)
estimates for PM2.5
33
Validation for 2004 (RO, SP, SE, CH)GAINS
(Background EMEP City-delta) estimates for
PM2.5
34
Projection for 2020 (RO, SP, SE, CH)GAINS
(Background EMEP City-delta) estimates for
PM2.5
35
Validation for 2004 (UK)GAINS (Background EMEP
City-delta) estimates for PM2.5
36
Projection for 2020 (UK)GAINS (Background EMEP
City-delta) estimates for PM2.5
37
Main uncertainties and sensitivities

• Definition of target domain for the urban
  increment (55, 1010, 1515 km)
• For health impact assessment more knowledge about
  the evidentiary studies required
• Critical for comparison with observations
• Uncertainties about urban PM2.5 emissions,
  especially from wood burning and industrial
  process emissions
• National input required
• Other sources (re-suspension)
• Treatment of low wind speed days
• More meteorological information (e.g., height of
  mixing layer) would be necessary
• Formation of secondary organic aerosols
• Validation by quality-controlled observations

38
Change in SOMO35 in response to setting urban NOx
emissions to zero Results of the CAMx model
39
Conclusions

• A methodology to estimate urban pollution at the
  European level has been developed and implemented
  for 473 cities
• Computed urban increments range up to 15 µg/m3
• The quantification of these urban increments is
  especially sensitive towards the definition of
  target domains and quality of emission data
• The computed urban increments explain a
  significant fraction of the difference between PM
  observations in cities and results of
  regional-scale models. However, uncertainties in
  the available monitoring data hamperan extended
  validation
• Ozone Further work is required to incorporate
  response to urban NOx emissions into GAINS