1. Michel Van Roozendael BIRA-IASB, Brussels, Belgium
Tropospheric NO2 from space: retrieval issues and perspectives for the future
Michel Van Roozendael
BIRA-IASB, Brussels, Belgium

2. Retrieval method (basics) Main issues regarding:
Overview
Retrieval method (basics)
Main issues regarding:
Spectral fitting
Stratospheric correction
Tropospheric AMFs
Cloud correction
How to assess the accuracy of our retrievals?
Challenges for the future

3. GOME tropospheric NO2 intercomparison
Why such differences?
Van Noije et al., ACP, 2006
Who is right?

4. NO2 remote sensing using DOAS
UV-Vis NO2 absorption is:
Structured
Independent of pressure
Weakly dependent on T°
Total atmospheric attenuation is small (<< 1)
 Atmospheric transmission follows Beer-Lambert law in a simple way:
SCDNO2 P

5. The 3 steps to tropospheric NO2 VCDs
STEP 1: DOAS  NO2 SCD
STEP 2:
Remove the stratospheric part  tropospheric NO2 (TSCD)
Strat. NO2
STEP 3:
Convert TSCD into tropospheric VCDNO2
NO2
Surface

6. STEP 1: Spectral fitting issues
Error on DOAS fit controlled by:
S/N ratio, limited by shot noise of detector
Possible systematic bias due to:
Temperature dependence of NO2 cross-sections
Interferences with unknown or badly known absorbers (e.g. absorption from water vapor and/or liquid water)
Inaccurate correction for Raman scattering by air and/or water
Instrumental artefacts. DOAS is insensitive to spectrally smooth radiometric errors, but very sensitive to “offset type” errors as well as to radiance errors displaying high frequency structures (e.g. polarisation, undersampling, …)
Choice of fitting interval  trade-off between S/N and minimisation of bias effects. Differences in settings/correction schemes applied by different groups may result in significant SCD differences.

7. Accuracy of measured radiances: what does matter for DOAS?
S/N ratio  the more photons the best (in practice trade-off between spatial/spectral resolution and S/N)
Instrument/radiometric calibration issues:
Wavelength calibration
Knowledge of instrumental slit function
Dark-current correction
Straylight correction
Polarization correction
Diffuser plate response

8. Examples of known instrumental problems affecting DOAS retrievals
GOME diffusor plate spectral features interfering with NO2 absorption  time-dependent bias, requiring special treatment
Richter & Wagner, 2001
Courtesy J. Gleason, NASA
OMI dark current mis-corrections leading to across-track fluctuations in the retrieved NO2 field  also requires the application of “soft calibration” procedures

9. STEP 2: Stratospheric correction
Different methods can be used to extract the tropospheric signal from the total column seen from space (e.g. use cloud shielding effect, limb-nadir matching, wavelength dependence of AMFs, etc)
By far, the most popular ones are:
The “reference sector” technique and its variants (e.g. harmonic analysis)  use NO2 columns measured over unpolluted regions to infer the stratospheric part over source regions
The model based technique  use NO2 columns from 3D-CTM constrained by observations over unpolluted regions
The assimilation technique  assimilate NO2 SCD in 3D-CTM (variant of model method)

10. STEP 3: get VCDs using tropospheric AMFs
Most complex and error prone part of the retrieval
Tropospheric NO2 AMFs depend on:
Solar and viewing geometries
Surface properties (albedo, ground elevation)
Aerosols
Cloud properties
Shape of tropospheric NO2 profiles
Problem: these properties are to a large extent unknown, or there are known at inappropriate resolution !

11. Examples of solutions currently in use
Property
Current treatment in AMF calculation
Groups
Surface albedo
- GOME/TOMS data base
All groups
Cloud fraction and cloud top height
- Screening based on cloud fraction
- Explicit correction using IPA and accounting for ghost column
- Bremen, Heid
- KNMI, NASA, SAO
NO2 profiles
- Scenarios
- Monthly mean profiles (MOZART)
- Daily profiles (GEOS-CHEM)
- Daily profiles (TM4)
- Heid, NASA
- Bremen
- SAO
- KNMI
Aerosols
- Neglected
- Scenarios (Lowtran)
- Implicitly corrected by cloud treatment
- Complex aerosol model
- Heid
- KNMI, NASA

12. Cloud correction scheme
Clouds shield surface NO2
Clouds enhance sensitivity to NO2 located above or at cloud altitude
Clouds generally treated as lambertian reflectors  effective cloud fraction and scattering cloud top height
AMF = (1-f).AMFclear + f.AMFcloud
AMFcloud requires estimation of the NO2 column underneath the cloud (ghost column) !
NO2 layer
Surface

13. Impact of clouds on tropospheric AMFs

14. How to assess the accuracy of our NO2 retrievals?
Differences in retrieval strategies result in inconsistencies beteween NO2 products derived from different groups. Problem even larger when different instruments are analysed by different groups.
Strategies to assess the accuracy of NO2 retrievals:
Comprehensive error analysis (cf. Boersma et al., 2004)
Intercomparison of satellite data sets (cf. van Noije et al., 2006)
Validation using external correlative data sets

15. Tropospheric NO2 validation: a challenge
Why is difficult to valide tropospheric NO2 from satellites?
NO2 emissions are extremely variable in space in time  the NO2 field as sampled by the satellite can hardly be matched by correlative measurements.
Suitable validation data sets are currently limited:
In-situ surface measurements (difficult to compare with satellite columns)
Remote-sensing network from NDACC (focus on stratospheric columns)
In-situ aircraft (excellent but expensive -> lack of statistics)
MAXDOAS (promising technique under development – need for network deployment)
NO2 Lidar (interesting but expensive -> lack of statistics)

16. Status of tropospheric NO2 sounders
Instrument
Satellite platform
Launch date
Equator crossing time
Resolution
Horizontal
Revisit Time
GOME
ERS-2
1995
10:30 LT
320x40 km2
3 days at equator
SCIAMACHY
ENVISAT
2002
10:00 LT
60x30 km2
6 days at equator
OMI
EOS AURA (A-train)
2004
13:30 LT
15x25 km2
1 day
GOME-2
METOP
2006
9:30 LT
80x40 km2

17. Current status: GOME, SCIAMACHY, GOME-2 and OMI
ERS2-GOME 10:30 LT 320x40 km2
SCIAMACHY 10:00 LT 60x30 km2
GOME-2 9:30 LT 80x40 km2
OMI 13:30 LT 15x25 km2

18. Requirements for future NO2 monitoring systems
Driving requirements for air quality (Capacity study)
Spatial resolution 5-20 km
Revisit time 0.5 – 2h
Can be met through:
Option 1: combination of (at least one) geostationary satellite and one sun-synchronous low earth orbit satellite (LEO)
Option 2: constellation of several instruments in LEO – a minimum of 3 instruments is needed to satisfy sampling requirements at mid-latitude
 Trade-off between Options 1 and 2 must be evaluated (ongoing CAMELOT study)

19. Challenges for the future (1)
How to ensure the consistency of the global NO2 observing system (GEOSS/GMES requirement) when the fleet of instruments expands more and more?
Evolve towards common retrieval approaches?
Rely on both operational (standardised) and scientific (state-of-art) retrieval approaches

20. Challenges for the future (2)
What to do to improve NO2 retrievals?
A) Enhance sensitivity to detect lower levels of pollution
Using better instruments  improve S/N ratio through better photon collection efficiency
Larger throughput (limited by weight and size!)
Longer integration time (GEO)
Multiply instruments
Using improved algorithms
Expand fitting range using direct-fitting  puts high requirements on the quality of Level 1 data, and on data processing

21. Challenges for the future (3)
B) Improve treatment of radiative transport
Use synergy with other (co-located) instruments to get better information on albedo, aerosols and clouds
Use more advanced model data or higher resolution
Improve cloud retrieval algorithms in synergy with those of NO2 (combined cloud-trace gas retrievals)
C) Get more than the column (vertical profiling)
Expand fitting range using direct-fitting and optimal estimation  requirements on Level 1 quality (cf. sensitivity)
Further develop cloud slicing techniques
Use dual/multiple view geometry?