1. Carbon monoxide from shortwave infrared measurements of TROPOMI: Algorithm, Product and Plans
Jochen Landgraf, Ilse Aben, Otto Hasekamp, Tobias Borsdorff, Haili  Hu, Joost aan de Brugh

2. The TROPOMI instrument
SWIR-3 channel
Assembled TROPOMI instrument
Because of innovative immerse grating technology , the SWIR-3 channel has only a volume of 5 liters.

3. The 2.3 μm spectral range Spectral range contains information on:
Water
HDO
Methane
Carbon monoxide
Quality requirement for operational data product:
CO: 10 % precision /
15 % accuracy
CH4: 1 % precision /
1 % bias

4. Level 1 data product (2:15 hr after sensing)
Data streams
Level 1 data product (2:15 hr after sensing)
Near real time NRT
Offline OFL
scientific
45 min after L1
2 weeks after NRT
SURFsara CO
CH4 CO
CH4,
CO,
HDO/H2O
Level 2 data product

5. The 2.3 μm spectral range
Quality requirement for operational data product:
CO: 10 % precision /
15 % accuracy
CH4: 1 % precision /
1 % bias
The design of the algorithms is driven by a trade-off between TROPOMI L2 data coverage and the required accuracy.
Different algorithms (SICOR for CO and RemoTeC for CH4)
Different data coverage (70 % versus 3 %)
Different accuracies

6. Physics based retrieval approach
Light path can be shortened or enhanced due to atmospheric scattering.
A physics based retrieval aims to infer information an trace gases and atmospheric scattering simultaneously from the measurement.
Both, the operational CH4 and CO retrieval are based on this principle, which makes the algorithm numerically demanding.
28/06/2015

7. CO data coverage
No useable signal for clear-sky scenes over ocean (except in glint geometry).
To improve coverage, we
have to consider cloudy observations as well, filtering on low altitude water clouds.
Mean SNR of SCIAMACHY 2.3 μm clear sky measurements

8. Operational CO algorithm: SICOR
Band 1: nm non-scattering retrieval, difference between retrieved CH4 and a priori knowledge used to filter on high and optically thick clouds Band 2: nm remaining atmospheric scattering is described by a triangular height distribution. Fit parameters: CO, H2O, scattering optical depth and scattering layer height, using a priori CH4. Retrieval uses a two stream RTM. Processing time: 0.15 seconds per ground pixel
Reference:
Gloudemans et al., 2009 / Vidot et al., 2012,
Borsdorff et al., 2014, 2015
Landgraf et al., 2016

9. SICOR: CO data product for cloudy atmosphere
Estimate retrieval performance from simulated measurements
CO mean bias: 1.6 %
Standard deviation: 2.3 %

10. Cloud retrieval from CH4 absorption features (1)
Two stream solver is one of the most simple RT solver accounting for multiple scattering but it introduces forward model errors to the retrieval.
So, the retrieved cloud parameters allow to describe the CH4 absorption well but introduces an bias for CO with a different vertical distribution

11. Cloud retrieval from CH4 absorption features (2)
Additionally errors are introduced by wrong CH4 a priori knowledge (XCH4 from TM5-NOAA of previous year, air mass and water from ECMWF forecast).
Overall 3 % CH4 uncertainty
3 % CO uncertainty depending on profile
Difference between CH4 total column dry air mixing ratios from TM5-NOAA simulations and GOSAT retrievals for the period June 2009 to December 2012.

12. TROPOMI orbit ensemble (1)
Orbit ensemble of simulated measurements with resampled MODIS cloud data to TROPOMI pixel size
cloud fraction
cloud optical depth
cloud top height

13. TROPOMI orbit ensemble (2)
MODIS surface albedo
Calypso cirrus optical depth

14. TROPOMI orbit ensemble – CO column
Biomass burning event central Africa
Overall very good precision
Small biases with some outliers <8% due to simple cloud model

15. CO Bias Probability Density Function
CO mean bias:
0.9 %
CO standard deviation: 1.1 %
Only very weak dependence of the CO bias on cloud occurrences
clear
sky
cloudy

16. SCIAMACHY heritage
10 years of SCIAMACHY CO total column, Borsdorff, T., et al., 2016 (in preparation)

17. SRON Plans Validation and algorithm improvement
TCCON, IAGOS for validation
NIR-SWIR combination (to use O2 A band instead of CH4 for cloud retrieval)
Combining cloudy and clear sky measurements to retrieve height information
TIR-SWIR combination
Small case studies to demonstrate the use of the CO data product in close collaboration with modelers.