Transcript TNO Presentation

On the estimation of emissions
from earth observation data
M. Schaap, R. Timmermans, H.
Denier van der Gon, H. Eskes, D.
Swart, P. Builtjes
Experiences from TNO using the LOTOS-EUROS model
TNO experience
TNO has a large experience in emission inventories
•
•
•
•
2
Emission inventories on global, European and national scale
Top-down and bottom-up
Much attention for spatial distribution
Delivered emissions to e.g. GEMS, MACC, EUCAARI,
MEGAPOLI
TNO experience
26 Nov 2009
Emissions of Elemental Carbon in Europe
How can satellite data help to improve these maps?
3
TNO experience
26 Nov 2009
Natural & Biogenic emissions – calculated online
Isoprene
Marine emissions
How can satellite data help to improve these algorithms?
4
TNO experience
26 Nov 2009
Research aims at TNO i.r.t. Earth Observation
• To combine earth observation data and modelling to obtain an
optimal assessment of the air quality over Europe.
• To quantify anthropogenic emission strengths by using EO data.
• Reanalysis as well as NRT
3
3
PM10 measurements (g/m )
5
10
15
20
25
30
35
40
Groundbased
5
TNO experience
45
AOT (-)
PM2.5 LOTOS-EUROS (g/m )
50
55
2
4
6
8
10
12
14
16
18
Model
20
22
24
26
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Satellite
26 Nov 2009
LOTOS-EUROS
•
•
•
•
CTM directed at the lower troposphere (up to 5 Km)
Developed at TNO & RIVM
Used at KNMI, PBL, Univ. Berlin, Univ. Aveiro
Includes a data assimilation environment
Numerical formulation
Input data
Emissions
Regional / Local
Preprocessing
Land use
Meteorological forecast
Explicit CTM
Transport
Advection
Turbulence
Chemistry
Gas phase
Aerosol
Deposition
climatology /
explicit model
Gases O3 , NO2 , SO2 …
Aerosols Sulfate, Nitrate,
sec. organic, primary…
ECMWF
Global chemical forcing
Gidded
hourly simulated
concentrations:
Wet and Dry
Wet, dry
deposition fluxes
Schematic of LOTOS-EUROS modelling system:
Meteorology
Emissons
Chemistry
Instantaneous
Emissions
Advection
Aerosol
physics
24Hr
Data
Land use
Vertical
exchange
Dry
Deposition
Satellite
data
Boundary
conditions
Wet
Deposition
…
…
Input
7
TNO experience
NO2
O3
PM
AOD
Chemistry transport model
Observations
EnKF
EnKF
filter
smoother
Data-assimilation
26 Nov 2009
Integration of earth observation data into models
Observations
Satellite data
Observations
In-situ data
8
TNO experience
Model input
A priory
State
Model
integration
Assimilation
Assimilation
Procedure
procedure
EnKF
Analysed State
Analysed
Concentrations
Emissions
State
(xa)
L4 data products
Emission estimates
Weather
forecast
Model
integration
Air Quality
forecast
Meteorology
Emissions
Noise
Output
Other parameters
26 Nov 2009
Single component assimilation – Ozone
Ozone measurements from the EMEP network assimilated
9
TNO experience
26 Nov 2009
Measured
Validation
Assimilation
station
Assimilation
200
Concentration
Vredepeel
Model
150
100
50
0
100
200
300
400
500
Hour
Model
Measured
600
Assimilation
Westmaas
Validation
station
Concentration
200
150
100
50
0
100
200
300
400
500
Hour
10
TNO experience
26 Nov 2009
600
Single component assimilation – PM10
Without assimilation
11
With assimilation
From Denby et al. 2008, Atmospheric environment 42
TNO experience
26 Nov 2009
The big challenges:
• To disentangle the uncertainty due to the emission input from
other model uncertainties
• The assimilation “blames” all errors to a limited amount of
parameters
• To keep the system realistic and balanced
• To combine different sources of data – multi component
12
TNO experience
26 Nov 2009
Assimilation of SO2 and SO4 – a case study
SO2
SO4
Annual mean for 2003
13
TNO experience
26 Nov 2009
Modelled annual mean concentrations
SO2 and SO4
SO2
14
TNO experience
SO4
OBS
1.6
2.5
MOD
2.3
1.8
OBS/
MOD
1.4
0.7
RESID
1.5
1.4
RMSE
2.1
2.1
Cor
0.48
0.47
26 Nov 2009
Results: SO2 annual cycle over all assimilation stations
15
TNO experience
26 Nov 2009
Results: SO2 & SO4 annual cycle over all stations by
including uncertain conversion rates
SO2
16
TNO experience
SO4
26 Nov 2009
Annual mean estimated multiplication factors
Emissions
Reaction rate
Also after acknowledging the shortcomings of the model it indicates
that the shipping emissions and those in Poland may be too high
17
TNO experience
26 Nov 2009
Assimilation OMI NO2 measurements with LOTOS-EUROS
Impact for ozone at the surface
Analysis NOx emissions / inventory (yellow=1)
18
TNO experience
26 Nov 2009
Impact of assimilation on ozone peak value
Assimilation
Surface ozone
No assimilation
Assimilation
OMI NO2
NO2 bias in the model effects
ozone negatively
Note, OMI NO2 may be ~25%
high
19
TNO experience
26 Nov 2009
Bias determination PM & AOD
PM10 = 2 * PM10model
Modeled AOD
To use AOD for estimating PM concentrations and
emissions
Daily average AOD over all stations
AODaeronet = 1.6 * AODmodel
Probability
AERONET AOD all data
20
TNO experience
26 Nov 2009
AERONET AOD - model
Assimilation strategy
• 2006
• LOTOS-EUROS CTM
• Bias corr.: AOD = AOD * 1.6
• Reduced domain
• MODIS data
• Uncertainty: 0.05 + 0.15AOD
• All pixels used
AERONET
EMEP
• EnKF Assimilation
• Model uncertainty relative to anthropogenic emissions
• 30%, Daily, time correlation of three days
• 12 ensemble members
• Model simulation at overpass time stored during the day
• Assimilation performed once a day at midnight
• Localisation (ρ = 50 Km)
21
TNO experience
26 Nov 2009
Model to MODIS comparison
MODIS composite
22
TNO experience
Modelled composite
26 Nov 2009
Effect of Assimilation
MODIS composite
23
TNO experience
After assimilation
26 Nov 2009
Impact of assimilation on comparison with MODIS
RMSE
Cor
24
TNO experience
26 Nov 2009
Verification with AERONET
Correlation
Assimilation
1
0
Model
Assimilation
RMSE
25
TNO experience
Model26 Nov 2009
1
Correlation
Assimilation
PM10 measurements
Model
Assimilation
RMSE
26
TNO experience
Model
26 Nov 2009
May 6th
May 7th
MODIS
Assimilation
Model
27
TNO experience
26 Nov 2009
Hamburg
2
1
28
TNO experience
26 Nov 2009
Neuglobsow
AOD
PM10
29
TNO experience
26 Nov 2009
Exploring the impact of forest fire emissions (FMI) on
calculated PM fields: May 6th, 2006
LOTOS-EUROS PM
0,9
0,8
0,7
0,6
LE
0,5
Hamburg
0,4
ForestFires
0,3
0,2
0,1
30
TNO experience
26 Nov 2009
0
1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
101
Conclusions
• To quantify emissions from observations one needs a model
• Data assimilation or inverse modelling of EO data is feasible
• We are able to provide level 4 products
• Data assimilation an objective framework
• To estimate emissions challenges are:
• To disentangle the uncertainty due to the emission input from
other model uncertainties
• To keep the system realistic and balanced
• To combine different sources of data – multi component
• Hence, this is a long term scientific research line
31
TNO experience
26 Nov 2009
Where can we use our present capabilities to provide
information on emissions?
• Search for trends in the parameter estimates?
• Does the EO data indicate that the emission trend is not as expected?
• The system does the meteo correction, etc for you.
• To indentify locations of new and significant emission sources
• The areas with consistently high model-measurement deviations
• To identify time profiles – needed: geostationary data
• Emission estimates
• Only in hotspot locations, and/or with observations during the emission
itself.
• Direct variables such as land use, LAI, Fire Radiative Power, White cap, etc
32
TNO experience
26 Nov 2009
Assimilation stations
33
TNO experience
26 Nov 2009
Validation stations
34
TNO experience
26 Nov 2009