GOME2 Error Study: Column Retrieval
Overview of Final Presentation (IUP Contribution):
Main contribution from IUP Radiative Transfer Modelling and Data Simulation Error analysis for GOME2 trace gas column retrieval
Contributors Rüdiger de Beek (retrieval, error analysis, RTM database) Vladimir and Alexei Rozanov (RTM development) Andreas Richter (DOAS settings, GOME1 error assessment) Marco Vountas (Ring effect) Mark Weber (project management, error analysis) John Burrows (PI University of Bremen, GOME1 lead scientist)
GOME2 Error Study: Column Retrieval
Topics: Task Report 1: Tool Adaptation and Definition of Data
DOAS trace gas column retrieval (WP120, M. Weber) RTM and spectral simulation of Input Datasets (WP130, M. Weber) Overview of error sources (WP150, M. Weber) Basic SNR error (WP 150, R. de Beek)
Task 2: Analysis of Error Sources Spatial Aliassing (WP 210, M. Weber) Spectral Resolution and Undersampling (WP 230, R. de Beek) RTM Assumptions and Earth Curvature (WP250, R. de Beek) BRDF (WP 260, M. Weber) Pointing and Geolocation (WP 270, R. de Beek)
Task 3: Optimal Operational Settings and Error Mitigation (M. Weber) Overall Error Budget Recommendations Future Work
WP120: Column Retrieval Technique
WP 120: Tools adaptation A) Trace gas column retrieval Retrieval Techniques:
DOAS = Differential Optical Absorption spectroscopy (Platt and Perner 1994) Basic assumptions:
weak trace gas absorption (10-4-10-2) negligible T-dependence of x-sections slow variation in Rayleigh- and aerosol scattering contribution with
condition fulfilled for NO2, OClO, and BrO, but O3 is a strong absorber!
Standard DOAS (Two step retrieval) slant column fit: linear fit to match superpositions of X-sections to observed differential optical
depth conversion to vertical columns via AMF calculation by RTM
Although modified DOAS (Diebel et al. 1996) or weighting function DOAS (Buchwitz et al., 2000) may be more appropriate for ozone, standard DOAS was used in this study for all trace gases as done in the current operational GOME1 retrieval
AMF error is dominated by a-priori assumptions (beyond scope of this assessment), here focus on slant column retrieval
WP 130: Acquisition of Input Data
Spectral fitting windows Recommendation based upon GOME1 experience O3 VIS as option investigated in selected cases
Analysis of data DOAS Algorithm: KVANT (Fortran 90, M. Eisinger) Linear mapping of errors (see WP150)
WP 130: Acquisition of Input Data
B) Simulation of GOME2 Spectra
Radiative Transfer Model Full spherical model CDIPI (Combined Differential Integration with Picard Iteration)
(Rozanov et al. 2001) Arbitrary viewing geometries (SZA<98°)
Limb Nadir
Spectral range: 240-2380 nm (SCIAMACHY range) IR: line-by-line, correlated-k Accuracy (UV/VIS): <1% (SZA<90°), <3% (SZA>90°)
Approximate spherical model CDI (Rozanov et al. 2000) Pseudo-spherical source functions, i.e. CDIPI w/o PI Accuracy: <2% in limb geometry above ~35 km tangent height Pseudo-spherical version as option (SCIATRAN/GOMETRAN compatible)
CDI sufficient for non-limb geometry as is the case for GOME2
WP 130: Acquisition of Input Data
Comparison between CDI and CDIPI
CDI sufficient for non-limb geometry as is the case for GOME2
Viewing and solar angles in a topocentric coordinate system (tcs) Satellite (SAT) Top of atmosphere (TOA) surface (GRD)
Plan-parallel atmospheres use only one fixed set of angles
Modifications to CDI during this study: BDRF (RPV formalism, ocean glint) Refraction Weighting function
WP 130: Acquisition of Input Data
Choices of origins of tcs: SAT, TOA, GRD
Snow BDRF 300-330nmSZA=40°
WP 130: Acquisition of Input Data
GOME2 spectral simulation Viewing geometries (ERS Propagator)
Jan, April, July, October Latitudes: 5N, 55N, 75N, 75S High (0.8) and low (0.05) albedo 24 trace gas scenarios
Scan Simulation zenith line-of-sight angles
- 46.5° to 46.5° Scan time = 4.5 sec. Sampling time = 0.01875 sec # of line-of-sights / = 240
IT=0.1875s 10 LOS (24 ground pixels per forward scan)
Table: 24 atmospheric scenarios
Month Latitude
deg SZA Deg
Rel. az. Deg
(east/west)
Albedo %
1 January 5 N 46.5 23.3/156.7 5 2 80 3 55 N 79.8 46.8/133.2 5 4 80 5 April 5 N 36.7 22.0/158.0 5 6 80 7 55 N 49.5 39.6/140.4 5 8 80 9 75 N 65.0 50.4/129.6 5
10 80 11 July 5 N 40.4 40.0/140.0 5 12 80 13 55 N 39.0 31.8/148.2 5 14 80 15 75 N 53.4 48.5/131.5 5 16 80 17 October 5 N 35.8 10.0/170.0 5 18 80 19 55 N 66.3 49.4/130.6 5 20 80 21 75 N 83.4 54.3/125.7 5 22 80 23 75 S 76.0 55.0/125.0 5 24 80
WP 130: Acquisition of Input Data
Realistic tracegas scenarios
SLIMCAT 3D CTM stratospheric profiles
Tropospheric modifications
Constant tropospheric O3 number density profile (all)
Biomass burning/ biogenic emission (5N, July) H2CO 2ppb < 3 km (may affect BrO fit) NO2 taken from MPI-2D CTM O3 doubled < 5 km
Free tropospheric BrO (55N, April) BrO 1ppt < 10 km (Fitzenberger et al., 2001)
PBL BrO explosion (75S, October, „ozone hole“) BrO 50ppt < 2 km O3 0ppm < 2 km
MPI 2D
Albedo dependent photochemical activity
BrO, April, 55°N
WP 130: Acquisition of Input Data
Comparison between MPI 2D (GOME1 tracegas climatology) to modified SLIMCAT
Ozone hole, 75S, Oct PBL low ozone event
high OClO, 75S, Oct chlorine activation
MPI 2D
MPI 2D as used in GOME1 V2.7 LV2 retrieval is outdatedGOME1 V3 climatology based upon TOMS V7 (O3) and US standard (NO2)
WP150: Overview of column errors
Overview of potential error sources
in trace gas column retrieval
diffuser plate spectral signaturediffuser plate spectral signature Spectral interference pattern
from sanded Al surface in solar irradiance
GOME2 currently uses the same diffuser plate
Errors of 50% and 70% in NO2 and BrO VC density, respectively
Without improving diffuser in GOME2 minor tracegas column retrieval not possible
O3 UV retrieval errors are ~0.3%
Richter and Wagner, 2001
WP150: Overview of column errors
Dichroic features in Channel 3Dichroic features in Channel 3 No specific investigation for GOME2 SCIAMACHY investigation on combined polarisation correction and dichroics
effect on O3 VIS retrieval reported in Appendix of Final Report
GOME1: no reliable O3 VIS retrieval & shift of NO2 fitting window to 425-450nm Different polarisation state measurements and polarisation correction scheme
in SCIAMACHY and GOME2 make a translation of the error to GOME2 difficult Dichroics are reduced in GOME2 as
compared to GOME1
Error in GOME2 polarisation correctionError in GOME2 polarisation correction Direct O3UV fitting of simulated error spectra
provided by SRON
O3 UV column error on the order of 0.3%
Richter and Wagner
WP150: Overview of column errors
Scan mirror degradationScan mirror degradation Different UV degradation rate for
irradiance and nadir spectra after 1999
No systematic trend in GOME1 V2.7 total ozone observed after 1999
Decrease in SNR due to blackening of the mirror effect column retrieval
DOAS retrieval is robust against instrumental degradation, indirect effect due to SNR changes, however, increases retrieval error
Bramstedt et al., 2002
Comparison of GOME1 total ozone with NH Dobson stations
WP150: Overview of column errors
Wavelength calibration errorWavelength calibration error Air-vacuum effect and outgassing
see dichroics radiometric calibration error (wavelength shift in key parameters)
Doppler shift in solar irradiance (0.008 nm) See undersampling error in WP230
DOAS retrieval does not require absolute radiometric calibration, however steep gradients in key parameter (dichroics) and noise due to interpolation (undersampling) error can introduce unwanted spectral artefacts
Wavelength shifts without secondary effects (see above) can be corrected using shift and squeeze to align radiances and x-section spectra
Dichroics are much reduced in GOME2 as compared to GOME1
WP150: Overview of column errors
Error sources investigated in Task 1 and Task 2Error sources investigated in Task 1 and Task 2 SNR (WP 150) baseline error/reference case Spatial Aliassing (WP 210) Spectral Resolution (WP 230)
Open Slit Defocussing Undersampling (interpolation) error
RTM assumption (WP 250) Refraction Pseudo-spherical approximations (SAT, TOA, GRD)
BRDF and ocean glint (WP 260) Geolocation and pointing error (WP 270)