Transcript Kein Folientitel

Monitoring of the GOME/ERS-2 Inflight Calibration
Parameters from GDP-4 Reprocessing
M. Coldewey-Egbers, S. Slijkhuis, B. Aberle, D. Loyola
Introduction
Wavelength Calibration
In 2006 an update of the GOME Level-0-1 processor (GDP-4) has been developed in order to reprocess the
entire data set. The main driver for this updated version was the new sun mean reference spectrum intensity
check, and the associated closing of the time gaps between sun mean reference spectrum updates on the
Level 1b product. This opportunity has been used to include other algorithm developments such as an
extension of the GOME on-fly calibration parameter database, and a slightly modified spectral calibration. For
the first time a fully homogeneous dataset is available that is used to monitor the instrument performance and
stability over its lifetime from 1995 to 2006.
In the framework of the ESA-project ’Long-Term Monitoring of GOME Calibration Parameters’ several
spectral emission lines of the PtCrNe hollow cathode lamp were identified to be improper for an exact
wavelength calibration, and therefore have been removed from the analysis. The lines did not meet the welldefined statistical criteria for all available lamp measurements.
Sun Mean Reference Spectra, PMD Signals and Q-factors
Figure 5 shows the standard deviation of the wavelengths of all emission lines for all available calibration
orbits between June 1995 and May 2003 for the old and the new calibration analysis. Largest changes can
be found at the beginning of channel 3, where three lines were excluded, and at the end of channel 4 around
760 nm, where the very unstable last line has been removed. The noise of the new wavelengths is much
smaller compared to the old calibration, except in channel 2, where only one line has been excluded.
Instrument degradation as well as the ERS-2 pointing problem since 2002 lead to a strong decrease in the
measured intensity of GOME spectral channels 1 and 2. Figure 1 shows the ratio of the sun mean reference
spectra from 1997 to 2006 to the corresponding reference spectrum from 9th January 1996. The intensity in
channel 1 is reduced by more than 90%. In channel 2 the decrease is still 40-50%, and in channel 3 it is 040%.
Figure 5:
Standard
deviation
of
the
wavelengths of all emission lines for
the old (open circles) and the new
(red dots) calibration. Filled black
dots denote the lines that were
removed from the analysis.
Figure 1:
Ratio of the sun mean reference spectra from 9th
January 1997 to 2006 to the corresponding
reference spectrum of 1996. Grey shaded areas
mark features caused by the dichroic filter, which
separates channels 3 and 4.
Intensity decrease:
90% at 240 nm and 50% at 325 nm
Figure 2 shows all sun mean reference spectra of GOME from July 1995 to June 2006 for four single
wavelengths (290nm - channel 1, 330nm - channel 2, 430nm – channel 3, and 760nm - channel 4). Black
curves denote uncorrected data. The low periodic variation is due to the seasonality of the sun-earth
distance, which is maximum in July and minimum in January. Large peaks in the time series for all
wavelengths at the beginning of 2001 are due to severe problems with the ERS-2 spacecraft. They can be
directly assigned to data gaps and GOME anomalies, such as instrument switchoffs, as regularly
documented in the GOME yearly anomaly reports (see http://earth.esa.int/ers/gome/performance/). Besides
the large peaks, several small peaks can be identified in the curves, which occur for different wavelengths
at different dates. They can be explained with etalon structures. The red curves denote the sun mean
reference data which are first corrected for the etalon effect Secondly, all spectra are normalised to 1 A.U.
(Astronomical Unit) in order to remove the seasonal dependence. Finally, they are normalised to the
intensity of the reference spectrum from 3rd July 1995 to calculate the percentage decrease. The intensity
decreased by 80% at 290nm and by 60% at 330nm until June 2006. The drecease in channel 3 (430 nm)
started in 2001 and reaches now 40%. In channel 4 at 760nm only minor changes are observed. A slight
decrease of 10% from 1995 to 2001, and then a short increase of 5% until 2006. The corresponding time
series for the three PMD signals are depicted in Fig. 3. The degradation of the PMD signals show almost
the same behaviour as for the corresponding wavelengths.
One of the key elements in the optical system of GOME is
the quartz predisperser prism. The refractive index of quartz
depends not only on the wavelength of the light passing
through it but also on the temperature of the prism. It is
expected, that the temperature increases along an orbit,
partly due to warming by the sun and partly because light
passes through the instrument. Those temperature changes
may affect the lamp measurements and therefore the
wavelength calibration. Figure 6 shows a correlation
between one single wavelength (759.96 nm) and the
temperature. However, this correlation is not existing in
channels 1 and 2. It is strongest in channels 3 and at the
end of channel 4.
Figure 6:
One single wavelength (759.96 nm,
black curve) and temperature at the
predisperser prism (red curve) as a
function of time.
Wavelength calibration:
Wavelengths are more stable now using GDP-4. In channels 3 and 4,
wavelengths correlate with the temperature measured at the predisperser prism.
Influence of the South Atlantic Anomaly on the Leakage Current
The four GOME detectors are random access linear photodiode arrays. One characteristic of these
devices is a certain amount of leakage current produced by thermal leakage. The leakage current is
monitored by periodically taken dark-side measurements. The South Atlantic Anomaly (SAA) is a region
with intense radiation in space near the Earth that causes damage to many spacecrafts in low Earth orbit.
The GOME measurements are affected by high-energy protons leading to large data spikes. For this
study, all GOME orbits crossing the SAA region during night time have been separated. Figure 7 shows
the leakage current in channel 4 for an integration time of 30 seconds for 10 consecutive orbits in 1997.
The third and fourth orbit from top crossed the SAA. Data are much noisier and contain large spikes.
Figure 2:
Sun mean reference
intensity
for
four
different
wavelengths
(from top to bottom: 290
nm, 325 nm, 502 nm,
and 639 nm) from June
1995 to June 2006. Red
curves are corrected for
etalon structures and
for 1A.U.
Figure 8 shows the noise of the leakage current measurements for 30s integration time and the year
1997. The noise level inside the SAA increases by a factor of two compared to the noise outside the SAA
region. The leakage current itself is slightly larger inside the SAA than outside the SAA (without figure),
that is due to the expected spikes on individual detector pixels. Calculation of the dark signal using these
measurements from inside the SAA may yield to a slight overestimation of the leakage, and therefore to
an underestimation of the real signal. The same analysis for the year 2000 and the other time patterns
confirms these results. The influence of the SAA on the darkcurrent and its noise level is largest for the
long integration times (e.g. 30 and 60 s). It becomes smaller for the shorter ones of 1.5 s.
Figure 3:
PMD Signals from
June 1995 to June
2006.
Sun mean reference intensity and PMD signals:
Large outliers and anomalies in 2001 can be explained with GOME switch-offs.
Low periodic variation is due to the seasonality of the sun-earth distance.
The so-called Q-factors are defined as relative correction factors
that transform the measured signal with fractional polarisation to
an unpolarised signal (see GOME, 2000). Figure 4 shows the time
series of all three Q-factors from June 1995 to June 2006. The
strong decrease of Q-factor 1 is connected to the different
degradation of the PMD 1 signal and the measured signal in
channel 2. The PMD decreases faster compared to the channel up
to the year 1999 and then from 2001 to 2006 the channel signal
decreases faster. Q-factor 2 increases slowly from 1995 to 2006,
that means the PMD signal is larger than the corresponding
channel, while the channel decreases faster, respectively (see
also Figs. 2 and 3). Q-factor 3 is more or less stable (0.15 to 0.2)
over the entire period. Measurements carried out during the
calibration of the GOME FM have shown that all three PMDs are
sensitive to light above 790 nm. Early in-flight solar data showed
that straylight appears to be worst in PMD 3 (13%), that explains
the initial non-zero Q-factor 3. The irregular large peaks and
outliers are due to GOME anomalies such as cooler switch-offs,
instrument or satellite switch-offs, on-board anomalies, or special
operations.
Blickrichtung
Figure 7:
Leakage current in channel 4
for 30s integration time and 10
consecutive orbits from 1997.
3rd and 4th orbit from top
cross the SAA region.
Figure 4:
GOME Q-Factors for each PMD
from June 1995 to June 2006.
Q-factors: Outliers and peaks due to cooler switch-offs, instrument and satellite
switch-offs, and special operations. Decrease and increase due to different
degradation of PMD and corresponding channel signal.
Figure 8:
Leakage current noise inside
(red) and outside (black) the
SAA region for the year 1997.
Leakage Current and South Atlantic Anomaly:
Leakage current measurements are noisier and contain large data spikes.
References
[1] GOME: Level 0 to 1 Algorithms Descriptions, Techn. Rep., DLR, ER-TN-DLR-GO-0022, 2000.
Cluster Angewandte Fernerkundung
DLR Oberpfaffenhofen
www.caf.dlr.de