Use of IASI data for tropospheric trace
gas retrieval
Current studies for new IASI products supported
by CNES
T. Phulpin (CNES), C. Clerbaux (IPSL/SA), P. Coheur (ULB), S. Turquety (IPSL/SA), C.
Camy Peyret (LPMAA), S. Payan (LPMAA) and P. Prunet (Noveltis).
GOALS
IT IS ESSENTIAL FOR CNES (French space agency) TO DEMONSTRATE
THE SOCIETAL BENEFITS OF SPACE PROJECTS LIKE IASI.
To foster use of IASI data for meteorology and other areas of
application (Climate, air quality monitoring, environment,
etc.) by supporting research, distributing high level products
and initiating new services.
TRYING TO IDENTIFY THE LIMITATIONS (with IASI) TO BE USED IN
DEFINITION OF NEXT GENERATION (PEPS)
Contribution and support to research
- Expand IASI sounding to cloudy pixels (Level1)
- Improve current level 2 products
- Study additional products
- Implement and distribute Level 3 or Level 4 products
METOP [2007-2020]
IASI
-
Inversion
O3, CO, CH4,
N2O,CO2,
HNO3,
SO2 (volcans),
CFCs
GOME2
algorithms
O3, NO2,
SO2, H2CO
+Declouding
Level 2
Concentrations
Global Averaged
assimilation
Tropospheric chemistry
Pollution transport, AQ
Emissions
Atmospheric
spectra
Level 3
CTM +
Data
Level 1
distributions
Level 4
Global distributions
(multisensors
PROCESSING HETEROGENEOUS SCENES
(B. Tournier, S. Bijac, O. Lezeaux, P. Prunet)
„ Use of heterogeneous pixels to increase the coverage
„Benefits :
Š Atmospheric profiles on the edges of frontal systems
Š Reducing integration time and spatial averaging domain for trace gases retrieval (CO2)
„ Issue of spectral shift and pseudo noise due to composition of different
spectra at different location within off axis pixels
(see P. Schlüssel Poster)
IASI heterogeneous scene algorithm
„ Two algorithms are designed
Š Scenes decomposition (studied for Eumetsat)
• Suitable for radiances direct assimilation.
• The selection of the homogeneous scene to be retrieved is allowed.
D and leads
• Retrieved scene type can be chosen, suitable for chosen cloudLtype
O
to cloud parameters retrieval.
• Amplification of the noise due to scene decomposition not controlled.
• Amplification of the noise due to spectral post calibration.
Š Direct model combination
• Suitable for full scene retrieval (all homogeneous components shallW
be
E
considered simultaneously). Needs accurate FRTM for CloudyNpixels and
knowledge of surface emissivity spectra
• No noise added
• System ready for any progress in direct models in the frame of clouds
processing.
These two approaches are currently under development and will be compared.
The most accurate and fast will be implemented for test on a routine basis
CO2
Mission specification : Requirements
„ Requirement on the IASI column weighted averaged
Boundary layer
Boundary layer
(pressure)
(height)
Weight with
weighting function
Weight with
weighting function
IASI/SPECTRE
IASI/DFT
1013-900 hPa
1.0 km
0.45
0.35
1013-720 hPa
2.7 km
0.77
0.7
Boundary layer
Boundary layer
(pressure)
(height)
Required accuracy
on mean content
(ppmv)
Required accuracy
on mean content
(ppmv)
IASI/SPECTRE
IASI/DFT
1013-900 hPa
1.0 km
0.9
0.7
1013-720 hPa
2.7 km
1.54
1.4
Required accuracy of about 1 ppmv on the IASI CO2 column weighted averaged
„ High spatial and temporal sampling of IASI
Spatial averaging : ~ 1000 measurement points for (103x103) km2
Temporal averaging : 60 measurements/month
5 % of clear data (20 % with cloud-clearing ?)
Theoretical noise reduction ratio of about 50
CO2
Data processing : Discrete Fourier Transform
(DFT) methodology
Selection of 16 specific spectral windows (1282
spectral samples): regions with strong
atmospheric CO2 absorption or emission
From the quasi periodic line structure of the IASI
spectrum
Š
Š
re-sampling on a periodic base built from the spectral
transitions of the CO2 lines
application of a Discrete Fourier Transform to specific
spectral windows
DFT pseudo data (mean, fundamental and first
harmonic)
M ea n sig n a l
F u n d a m e n ta l
F ir st
H a r m o n ic
TEMPORARY CONCLUSIONS
CO2
DFT approach
• Permits an efficient extraction of the CO2 information
– Filtering the impact of other variables (to be confirmed for surface parameters)
– No significant loss of information (to be confirmed near the surface)
– Data compression
• Gives reliable results on measured noisy data, consistent with independent estimates, showing
the robustness of the algorithm
Validation on representative sets of data
Retrieval accuracy from a single spectrum
•
•
IASI Balloon
– 4 ppmv (1 %) on the column averaged mixing ratio
– no reliable estimate in low troposphere
IASI Metop
– 1.2 ppmv (0.3 %) on the column averaged mixing ratio
– 20 ppmv (5 %) in low troposphere: information correlated with free troposphere
CO2 information from IASI is at the level of mission specifications for a single spectrum
Retrievals from simulated data indicate that IASI CO2 product on 3 layers would be more efficient than
a column average, in order to exploit the low level information present in the data
Remaining to be done :Explore the potential of spatial and temporal averaging
Explore the CO2 processing in partially cloudy scenes, in order to separate upper and lower
troposphere information
THIS GOING TO BE STUDIED IN THE NEXT MONTHS
Aerosols
„ Infrared spectra are sensitive to aerosols specially absorbing aerosols like sand
dust, volcanic ash and biomass burning
„ Important parameters which can be retrieved from spectra are optical depth and
altitude (Pierangelo, 2005)
„ Algorithm developed at LMD and applied to AIRS has been expanded to IASI. A
selection of IASI channels dedicated to aerosol has been proposed.
TRACE GAS RETRIEVAL FROM IASI
„ Level 2 Processing based on Neural network at stage 1.
„ Those networks have been trained using pseudo IASI spectra
simulated from IMG data
„ Alternative inversion method is 1d Var. This method has been used
to analyze sensibility of retrievals to various parameters. It will be
implemented in a research mode and tested to compare its results
with operational retrievals
IMG distributions using IASI processing tools (operational mode)
Ts
O3
CH4
CO
+ Cloud filtering
Hadji-Lazaro et al., GRL
2001
Turquety et al. GRL 2002
Clerbaux et al., ACP 2003
Clerbaux et al., IEEE 1999; JGR 2001
Hadji-Lazaro et al., JGR 1999
SA-Neural network [Turquety et al, JGR 2004] implemented in
the Eumetsat ground segment for IASI real-time data
processing
OZONE
Errors
Partial columns
Vertical profiles
12-24 km
Ground-12 km
a priori variability
total error
40
[O3]retrieved (DU)
error sources:
smoothing
measurement
upper-air temperature profile
ILS
30
altitude (km)
100
20
Ny Alesund
Uccle
Hilo
Wallops
Easter Island
Lauder
Neumayer
300
75
200
50
50
R = 0.75
bias = 21 %
σ = 36 %
25
0
25
50
75
error (%)
100
125
150
100
75
10
0
24-30 km
25
50
75
100
100
R = 0.98
bias = -3 %
σ = 10 %
100
200
300
R = 0.93
bias = -1 %
σ = 18 %
25
25
50
75
100
[O 3]sondes (DU)
Accuracy and precision:
► High for the lower and midlle stratospheric columns
► Good for the tropospheric columns in most cases (bias decreases to 4% partial columns < 25
DU are neglected)
[Coheur et al. JGR 2005 ]
CARBON MONOXIDE
Boundary layer CO (1 km)
Upper tropospheric CO (10 km)
Pollution
Transport
Biomass burning
Convective transport
Good agreement with surface measurements (CMDL network) and
model distributions (GEOS-CHEM)
[Barret et al. ACPD 2005 ]
Nitric Acid (HNO3)
Volcanic plumes (SO2)
Tb =f (colonne SO2)
Sensitivity to level and
distribution of SO2
300
1166,5
1168
1367
1370
HIRS ch 12
1166,5 T
1168 T
1367 T
1370 T
HIRS 12 T
290
280
270
260
Much stronger at 1350 cm-1
250
If strong amounts a combination of
the two bands could be used to
retrieve Integrated amount and
plume altitude
240
230
220
210
200
0,00E+00
5,00E+02
1,00E+03
1,50E+03
2,00E+03
2,50E+03
3,00E+03
Ecarts avec profil 6
5
0,5 320
320
315
310
310
0
2,5
305
300
300
SO2 *1000
295
290
290
SO2 *1000 9
285
0
-0,5
SO2 500 8 et 9
-2,5
SO2 *250 500 250
-1
280
-5
Tb (K)
275
280
-1,5 270
270
-7,5
265
260
260
-2
255
-10
250
250
-2,5
245
-12
SO2 *E04
240
240
Delta 8-9 5000 5000
235
-3
230
delta 9 10000
230
-15
delta 10 2500 5000 2500
225
Wavenumber (cm-1)
1500
1480
1460
1440
1420
1400
1380
1360
1340
1320
1300
1280
1260
1240
1220
1200
1180
1160
1140
1120
1100
1080
1060
1040
1020
1500
1480
1460
1440
1420
1400
1380
1360
1340
1320
1300
1280
1260
1240
1220
1200
1180
1160
1140
1120
1100
1080
1060
1040
1020
1000
-17
1000
-3,5 220
220
OZONE FROM IR-UV
Vertical sensitivity
GOME AVK
IMG AVK
30 km
Altitude (km)
30
20 km
18 km
16 km
14 km
12km
20
10
10 km
1 km
0
-0 .1
0 .0
0 .1
a ve ra g in g ke rn e l
0 .2
0 .3
OZONE PROFILE FROM IR AND UVS SPECTROMETERS
(P. Coheur et al.)
Taking advantage of multiple instruments
` Synergetic retrievals
O3 profile retrieval from IR and
UV measurements
A GOME → IMG
A GOME
40
30
ds
20
UV
3.4
IR
5.4
UV-IR
7.7
10
10
20
30
40
10
20
30
40
Altitude (km)
-0.20
-0.12
-0.04
0.04
0.12
0.20
0.28
` Application to GOME-2 / IASI
Solène Turquety, Roeland Van Oss
1/3
CARBON MONOXIDE
ASSIMILATION OF CO DERIVED FROM IASI IN A CTM
(C. Clerbaux and A. Klonecki)
OBSERVATION OPERATOR BASED ON MOPITT DATA
STRATEGY
„DEVELOPEMENT AND TESTS IN LABS BASED FIRST ON SIMULATED
DATA AND THEN ON DATA SAMPLES FROM UMARF
„IMPLEMENTATION IN ROUTINE PSEUDO OPERATIONAL
PROCESSING
Š AT CMS FOR LOCAL
Š AT METEO France TOULOUSE FOR GLOBAL
Š OR ON ETHER (National Data Center for Atmospheric chemistry) FACILITIES
„COMPARISON WITH OTHER GROUPS IN THE FRAMEWORK OF THE
ISSWG
„EITHER IMPLEMENTATION IN THE CGS, THE SAFs OR DISTRIBUTION
BY ETHER
Conclusions
Besides its capacity for atmospheric profiles through assimilation
in NWP models,
POTENTIAL APPLICATIONS OF IASI ARE NUMEROUS AND MANY
ADDITIONAL PRODUCTS ARE TO BE DEVELOPED
According to simulations, IASI looks then very promising for
greenhouse gases monitoring.
Studies are continuing and new topics like cloud properties
retrieval, surface parameters are being considered
About one year after MetOp launch CNES and Eumetsat organize a Joint
Workshop (study conference?) on the first IASI results which will initiate
ISSWG phase 2. We hope to see you there!!!