GOME/ERS-2 » Radiance Degradation Correction

Comparison of the solar irradiance spectra measured by GOME through the lifetime of the sensor with early GOME solar irradiance spectra or other space instruments, showed that the pre-flight radiance parameters were no longer applicable to the GOME in-flight situation [Peeters et al. (1996), Eisinger et al. (1996), Peeters et al. (1997)]. The GOME sensor shows degradation in all wavelength regions due to damages in its optical path. Most damage is done by hard UV light to the scan mirror [Snel (2000)], but there might also be damage to the quartz-glass of prisms and lenses, to coatings, or to the detectors. Several changes in sensitivity of the instrument are simultaneously monitored: degradation, geometric changes in optical paths, changes of coatings.


The correction algorithm for GOME science channel degradation was initiated at DLR in the framework of the Project GDAQI (GOME Data Quality Improvement). The approach chosen is the comparison of all GOME solar irradiance spectra with GOME solar irradiance spectra of July 3, 1995. This study was done for GOME science channels and PMDs.
The changes have been determined by building ratios of all solar spectra with the solar spectrum of July 3, 1995. A correction for Sun-Earth distance was applied to all solar spectra:

$C_{Sun Earth Distance}(t)$

Also seasonal variations have been identified:


This approach is independent from the absolute errors in radiometric calibration of the GOME spectra and only shows the time dependent changes in the GOME sensitivity.
$\displaystyle \frac{I_{sun}(\lambda, t)}{I_{sun}(\lambda,t=\mathrm{July\,3rd\,95})}$ $\textstyle =$ $\displaystyle P_{Degradation}(\lambda, t)
\cdot C_{Sun Earth Distance}(t)$  
    $\displaystyle {}\cdot C_{BSDF_{az}}(t) \cdot \mbox{Residual}(\lambda, t)$ (1)

The ratio of GOME solar measurements obtained in July 1996 to 2005 with a reference measurement in July 1995 (Figure 1) shows several structures of high and low frequency. The high frequency residual is due to:

• changes in the etalon structure in all channels
• shifts in the overlap regions, changes in dichroic structures in channel 3 and 4
• changes in the light source, the sun mainly in channel 1.

\begin{figure}\epsfxsize =\textwidth \epsffile{degrad_after_8_years.ps}\end{figure}

Figure 1: Degradation of GOME science channels in 15 years.

The low frequency structure can be described with polynomials for each channel. The spectra are fitted by polynomials in wavelength, and subsequently each wavelength is fitted by a polynomial in time. Although there is evidence to use exponential functions of time to model degradation the other phenomena monitored have different behaviour.
$\displaystyle P_{Degradation}(\lambda, t)$ $\textstyle =$ $\displaystyle \sum^n_{i=0} a_i(t)\cdot(\lambda-\lambda_0)^i$ (2)
$\displaystyle a_i(t)$ $\textstyle =$ $\displaystyle \sum^m_{j=0} b_{ij}\cdot(t-t_0)^j$ (3)

A constant solar output since 1995 is assumed. Because the solar activity since launch was low, it is sufficient to exclude solar lines that are influenced strongly by changes in solar activity (See [Weber et al. (1998)]) from the retrieval. The broadband wavelength dependent degradation is modeled using polynomials with low degree. For Channel 1 and 2 a second order polynomial is used in the wavelength domain. For Channel 3 and 4 only an offset can be fitted due to interference of dichroic shifts and low etalon frequency. The parameters $a_i$ show smooth behavior in time and therefor can be modeled with analytical functions in time.For more details see [GDAQI final report (2000)].

Correction of Seasonal Variations

Also seasonal variations due to insufficient characterization of the GOME calibration unit, affecting all channels of solar measurements have been identified. The correction algorithm was performed at DLR in the framework of the Project GOME SUPPORT.
The variations are most pronounced (6%) for wavelengths below 260 nm This variation is due to insufficient correction of the solar azimuth dependency


of GOME calibration unit. lv1 product is corrected for azimuth dependency using following formula:
$\displaystyle BSDF_{az}(\phi_{az})$ $\textstyle =$ $\displaystyle 1 - a \phi_{az}^2$ (4)

Thus there is a different behaviour for positive and negative $\phi_{az}$ the extractor software provides an option to remove this correction (parameter $a$ is part of lv1 product) and to use a new correction including an additional linear parameter:
$\displaystyle BSDF_az(\phi_{az})$ $\textstyle =$ $\displaystyle 1 + a_1 \phi_{az} - a_0 \phi_{az}^2$ (5)

New Algorithm Updates

The algorithms described above were used to generate degradation files with version 1.x. Since then, the following improvements have been made:
  • Version 2.x
The second-order polynomial fit for asymmetric BSDF has now been replaced with a look-up table which describes the azimuth dependence of the BSDF with greater accuracy (see [Slijkhuis (2004)])
  • Version 3.0
A new correction for degradation of the reflectivity has been derived in the framework of the CHEOPS-GOME project (see [Krijger (2005)]). This incorporates a correction for the differential degradation of the Solar and Nadir lightpaths, as well as an implicit recalibration of the diffuser BSDF keydata (applied to the Nadir reflectance, not to the Solar spectra, see [Slijkhuis (2006)])
  • Version 4.x
The degradation correction has been significantly improved, especially for channel 1. The time dependence is now described by a lookup table, instead of by a single polynomial. The wavelength depencence of the degradation is still fitted with a polynomial, but for the time dependence a low-pass filtering using piece-wise polynomials has been employed.

Residual Structures of Degradation Correction

Using all available solar measurements until 09-OCT-2007, degradation parameters as given in degradation file scdegrad.400 have been determined. The deviations between the measured and parameterized solar degradation corrections are given in figures 2, 3, 4 and 5. For these plots the BSDF azimuth parameters are used as given in the same file.
\begin{figure}\epsfxsize =\textwidth \epsfbox{/home/aerosol14/degradation/data/output/version_109/residual_K1.ps}\end{figure}

Figure 2: Residuals in Channel 1

Channel 1: Ratios of spectra with July 3, 1995 corrected for degradation and asymmetric BSDF. Etalon changes compared with July 3, 1995 dominate the residuals at higher wavelengths. For solar measurements since July 1998, the correction for low wavelengths (up to 260 nm), does not work accurately. This is due to changes in solar activity and due to undercorrected changes in scan mirror degradation. Improved algorithms to account for these changes are under development.
\begin{figure}\epsfxsize =\textwidth \epsfbox{/home/aerosol14/degradation/data/output/version_109/residual_K2.ps}\end{figure}

Figure 3: Residuals in Channel 2:

Channel 2: Ratios of spectra with July 3, 1995 corrected for degradation and asymmetric BSDF. Etalon changes compared with July 3, 1995 dominate the residuals.
\begin{figure}\epsfxsize =\textwidth \epsfbox{/home/aerosol14/degradation/data/output/version_109/residual_K3.ps}\end{figure}

Figure 4: Residuals in Channel 3:

Channel 3: Ratios of spectra with July 3, 1995corrected for degradation and asymmetric BSDF. Etalon changes and features of the dichroic mirror compared with July 3, 1995 dominate the residuals. In the overlap region 3-4 ($\lambda$ > 590 nm) degradation correction does not work well because of shifts in the dichroic mirror.

\begin{figure}\epsfxsize =\textwidth \epsfbox{/home/aerosol14/degradation/data/output/version_109/residual_K4.ps}\end{figure}

Figure 5: Residuals in Channel 4:

Channel 4: Ratios of spectra with July 3, 1995corrected for degradation and asymmetric BSDF. Etalon changes compared with July 3, 1995 dominate the residuals. In the overlap region 3-4 ($\lambda$ < 610 nm) degradation correction does not work well because of shifts in the dichroic mirror.
It can be concluded that the degradation of solar measurements is well parameterized with polynomial functions of low degree for wavelengts above 260 nm.


The degradation algorithm is implemented into the extraction software gdp01_ex. An option to apply the science channel degradation correction to sun, moon and earthshine spectra is available since version 2.0. The new BSDF-correction is also implemented. The default is not to corrected for degradation and for asymmetric BSDF azimuth behaviour (BSDF from on-ground calibration is used). For the optional Degradation- and (improved) BSDF-correction, a parameter file will be read by the extractor (e.g. scdegrad.302 ).
Please note the following:
  • New level 1 file format and new degradation file format for gdp01_ex v4.0
The polynomial coefficients over time, for degradation correction, have now been replaced by a look-up table. The level 1 data product generated by the operational GDP 4.0 software underwent a format change. Version 4.0 of gdp01_ex neither works with old level 1 files, nor with the old degradation files. Since the accuracy of the degradation correction is now independent of the goodness of a polynomial fit, which was better for short timescales, there is now no need anymore to retain degradation files which are optimised for the beginning of the GOME mission. Therefore, version 2.x/3.x files were not mapped to version 4.
  • An additional reflectivity degradation correction may be applied in gdp01_ex v3.02
Reflectivity data are present on the Degradation file versions 3.02 and higher.
  • New file format for gdp01_ex v2.3
The second-order polynomial fit for asymmetric BSDF has now been replaced by a look-up table, for greater accuracy. Version 2.3 of gdp01_ex does not work anymore with the old degradation files. These are now replaced by newly formatted versions. These contain the new BSDF correction, with the degradation data copied from older files. Degradation file versions 1.x are mapped to corresponding versions 2.x .
  • the option '-e <file_name>' in the version v2.0 of the gdp01_ex
does not switch on the asymmetric BSDF-correction for solar measurements as specified in the (older verions of the) user manual and help output. Both '-e <file_name>' and '-f <file_name>' flag should be used to correct for degradation and for asymmetric BSDF.

Degradation Files

Degradation parameters are delivered approximately every 6 months. For accuracy reasons, validity is dated back from last solar measurement used for retrieval.
ATTENTION: Due to a lack of sun measurements from March 2004 onwards, the degradation algorithm does not work very well with GDP version 2.x data beyond 2004. This problem has been removed with the reprocessing using GDP 3.02. For channel 1, another issue is that the parametrisation of degradation using a polynomial in time is becoming more and more inaccurate as degradation increases. For data before June 2001, the channel 1 degradation polynomials from file scdegrad.209 (and earlier) are in fact more accurate than those from later 2.x/3.x file versions. The degradation lookup table, introduced with GDP L01 version 4.x, has removed this issue.


Valid from

Valid to

Last Sun used

4.30 15-JUN-199504-JUL-201102-JuL-2011
4.20 15-JUN-199516-JAN-201116-JAN-2011
4.10 15-JUN-199517-OCT-200917-OCT-2009
4.00 15-JUN-199509-OCT-200709-OCT-2007
3.02 15-JUN-199522-JUL-200622-JUL-2006
2.13 15-JUN-199520-APR-200514-MAR-2005
2.12 15-JUN-199507-SEP-200423-JAN-2004
2.11 15-JUN-199503-JUN-200304-MAY-2003
2.10 15-JUN-199501-FEB-200309-FEB-2003
2.09 15-JUN-199507-JUN-200107-JUN-2001
2.08 15-JUN-199510-APR-200122-APR-2001
2.07 15-JUN-199505-JAN-200117-JAN-2001
2.06 15-JUN-199520-SEP-200004-OCT-2000
2.05 15-JUN-199515-JUL-200029-JUL-2000
2.04 15-JUN-199531-MAR-200009-APR-2000
2.03 15-JUN-199531-DEC-199926-JAN-2000
2.01 15-JUN-199530-JUN-199930-JUN-1999
2.00 15-JUN-199530-JUN-199930-JUN-1999

Reference Documents


The algorithm for degradation correction was first developed at DLR by Ernst Hegels in the framework of GDAQI (GOME Data Quality Improvement), an ESA/ESRIN funded project. The asymmetric BSDF correction was developed at DLR in the framework of GOME Support; the tabulated BSDF functions and the reflectivity correction were developed at DLR and SRON respectively, in the framework of the CHEOPS-GOME study (Climatology of Height-resolved Earth Ozone and Profiling Systems for GOME): both projects also funded by ESA/ESRIN. The co-workers in the GDAQI project, Ilse Aben, Cristina Tanzi (SRON) and Michael Eisinger (ESA/ESTEC), and further co-workers in the CHEOPS-GOME project, Matthijs Krijger, Jochen Landgraf (SRON) are acknowledged for their contributions. The degradation files were generated initially by Ernst Hegels, then by Sandra Wahl and more recently by Peter Hoffmann and Melanie Coldewey-Egbers.

For further information please contact:

Sander Slijkhuis, DLR-IMF
E-Mail: Sander Slijkhuis [at] dlr.de

Melanie Coldewey-Egbers, DLR-IMF
E-Mail: Melanie.Coldewey-Egbers [at] dlr.de

DLR IMF AP Copyright 2010 | Imprint