Journal of Atmospheric Chemistry (2005) 50: 295–320
DOI: 10.1007/s10874-005-5544-1
C
Springer 2005
On the Retrieval of Volcanic Sulfur Dioxide
Emissions from GOME Backscatter Measurements
W. THOMAS1,∗ , T. ERBERTSEDER2 , T. RUPPERT2 , M. VAN ROOZENDAEL3 ,
J. VERDEBOUT4 , D. BALIS5 , C. MELETI5 and C. ZEREFOS6
1
Deutsches Zentrum für Luft- und Raumfahrt (DLR), Institut für Methodik der Fernerkundung
(IMF), Oberpfaffenhofen, Germany, e-mail: [email protected]
2
Deutsches Zentrum für Luft- und Raumfahrt (DLR), Deutsches Fernerkundungsdatenzentrum
(DFD), Oberpfaffenhofen, Germany, e-mail: [email protected], [email protected]
3
Institut d’Aéronomie Spatiale de Belgique (BIRA-IASB), Bruxelles, Belgium,
e-mail: [email protected]
4
European Commission – Joint Research Centre, Institute for Environment and Sustainability (IES),
Ispra (VA), Italy, e-mail: [email protected]
5
Aristotle University of Thessaloniki, Laboratory of Atmospheric Physics, Thessaloniki, Greece,
e-mail: [email protected], [email protected]
6
Laboratory of Climatology and Atmospheric Environment, Athens, Greece and Foundation for
Biomedical Research of the Academy of Athens, Greece, e-mail: [email protected]
*Present address: Deutscher Wetterdienst (DWD), Department of Climate and Environment, 63067
Offenbach, Germany
(Received: 27 April 2004; accepted: 27 October 2004)
Abstract. We focus on the retrieval of volcanic sulfur dioxide (SO2 ) emissions from an analysis of
atmospheric UV backscatter spectra obtained by the Global Ozone Monitoring Experiment (GOME)
spectrometer on board the ESA European Remote Sensing Satellite (ERS-2). Here, the last major
eruptions of Mt. Etna on Sicily (Italy) in July/August 2001 and October/November 2002 provided an
excellent opportunity to study the retrieval of SO2 columnar amounts from ground-based, LIDAR and
satellite measurements. Our study shows that the bulk of emitted SO2 was confined in the troposphere,
mainly between 700 hPa and 400 hPa which is confirmed by trajectory analysis, by LIDAR observations and AVHRR observations. The area of influence of Mt. Etna eruptions ranges from the Western
Saharan Desert to Greece and the near east states and even down to the basin of Tschad, Africa. Our
analysis revealed that information about the plume height of volcanic eruptions and aerosol parameters is necessary for a reliable quantitative retrieval of SO2 from space-borne sensor data at periods
perturbed by volcanic eruptions.
Key words: GOME measurements, sulfur dioxide retrieval, trace gases, volcanic emissions
1. Introduction
Sulfur dioxide (SO2 ) emissions are of both natural and anthropogenic nature. A
major natural contribution originates from volcanic activity (Graf et al., 1997)
296
W. THOMAS ET AL.
while man-made emission is mainly due to the combustion of fossil fuels. In the
stratosphere, SO2 is depleted within weeks, whereas the lifetime of tropospheric
SO2 is counted in days. High tropospheric levels of SO2 are therefore usually
measured not very far away from their source (Eisinger and Burrows, 1998; Zerefos et al., 2000). Most emission sources are near the ground, the exceptions are
explosive volcanic eruptions such as the Mt. Pinatubo eruption in 1991 where a
large amount of SO2 was transported into the lower stratosphere (Hansen et al.,
1992).
A relatively silent phase of Mt. Etna on Sicily (Italy) finished in November
1999 and since then, several major outbreaks of Europe’s largest and most active
volcano were registered. The eruptions in July/August 2001 and the more intense
outbreak in October/November 2002 were observed by different satellite sensors.
The observations provided detailed visual information of the spatial distribution of
smoke and ash clouds, but in the case of the GOME spectrometer on board ERS-2
also of volcanic SO2 emissions into the atmosphere.
First SO2 column measurements from space using the Total Ozone Mapping Spectrometer (TOMS) were presented by Krueger (1983). McPeters (1993)
used the NOAA-11 Solar Backscatter Ultraviolet (SBUV/2) instrument to analyze the SO2 emission of the eruption of Pinatubo. Limitations of SO2 measurements using the TOMS sensor are discussed in detail by Krueger et al.
(1995). First examples of GOME’s capability to measure total atmospheric
SO2 columns after volcanic eruptions were shown by Eisinger and Burrows
(1998).
In our paper we focus on the retrieval of the total vertical content of atmospheric
SO2 from GOME backscatter measurements in the UV spectral range during and
after volcanic eruptions. In general, the retrieval of volcanic SO2 emissions from
space suffers from the sparse temporal and spatial coverage of actual satellite sensors
but also from clouds in the troposphere. The short tropospheric lifetime of SO2 and
the separation of the background SO2 content from volcanic emissions also hampers
the estimation of the volcanic source strength.
Other typically unknown but important parameters are the aerosol loading in
and around the volcanic plume and the height above ground level of emissions
(particles and gases). A sensitivity analysis is performed to identify the main physical parameters that influence the retrieval results. We show that the combination of trajectory analysis, ground-based measurements and GOME observations
can be used to confirm the presence of large amounts of SO2 of volcanic origin away from the volcano. For the first time, GOME-derived SO2 total columns
were successfully validated against ground-based measurements from a Brewer
spectrophotometer.
The GOME instrument and the data processing is described in Section 2,
slant column results and a discussion of radiative transfer calculations is given
in Sections 3 and 4, while total column results and validation results are presented
in Section 5. Our conclusion is given in Section 6.
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
297
2. GOME Instrument and Data Processing
The GOME spectrometer, an atmospheric chemistry instrument on board ESA’s
ERS-2, is able to measure the content of a number of minor atmospheric trace constituents including SO2 . GOME is a nadir-looking across-track scanning instrument
with a typical footprint size of about 320 × 40 km2 . It measures the back-scattered
radiation from the earth-atmosphere system between 240 nm and 790 nm with a
moderately high spectral resolution of about 0.2 nm to 0.4 nm. In addition, a sun
spectrum is recorded every day via a diffuser plate. More details about the instrument and first results can be found in Burrows et al. (1999b). In our study we used
an enhanced version of the GOME Data Processor (GDP) for the SO2 retrieval as
briefly described in Thomas et al. (2003).
The GOME ground pixels cover a relatively large area and a preprocessing step
to determine the cloud coverage of a footprint is required. Clouds are opaque in
the measurement range of GOME (except optically thin cirrus clouds), and this
needs to be taken into account if trace gas total columns are retrieved. We analyze
GOME backscatter measurements in and around the oxygen A-band between 758
nm and 778 nm following the method presented by Kuze and Chance (1994). Cloud
coverage results from GOME tend to be lower than corresponding results from other
satellite sensors (Koelemeijer and Stammes, 1999), but we focus on scenes with
low cloudiness and the impact of cloud retrieval results remains well below other
possible error sources.
The trace gas retrieval is based on a DOAS (Differential Optical Absorption
Spectroscopy) fitting technique that provides trace gas slant column amounts along
the viewing path of the instrument (Burrows et al., 1999b). The DOAS technique
involves a multilinear regression of GOME-measured optical densities against a
number of reference spectra. The (spectral-dependent) trace species path absorption is modelled according to Beer’s law (Equation (1)); when there are several
absorbers, the contributions are additive provided that the total optical density remains small and saturation does not occur.
dI(λ) = −I (λ) · σ (λ) · C(s) ds
(1)
The incremental change dI(λ) of the incident radiation I (λ) is proportional to the
cross-section σ (λ) of the trace species of interest, and the trace gas column amount
C(s) along the slant path ds in the atmosphere. For application to GOME backscatter
data, Equation (1) is integrated, logarithms of sun-normalized intensities are taken
and a low-order polynomial is added to account for broad-scale molecular and
aerosol scattering and also the reflection from the Earth’s surface:
n
I (λ)
log
=−
SCi σi (λ) − P(λ)
I0 (λ)
i=1
(2)
where SCi are the slant column densities of each species, σi (λ) the corresponding
cross-sections, and P(λ) the polynomial.
W. THOMAS ET AL.
298
The SO2 retrieval was performed between 315.8 nm and 327 nm (around 120
spectral points), which is slightly different from the window used by Eisinger and
Burrows (1998) but optimized for minimum slant column fit residuals. Nitrogen
dioxide and especially ozone are interfering species in the given spectral window
and are fitted simultaneously. Reference spectra at 221 K and 241 K were taken
from Burrows et al. (1998) and Burrows et al. (1999a), respectively. A correction
for the so-called solar Io -effect was applied to the ozone reference data following
Aliwell et al. (2002). This correction is vital since Beer’s law presented in Equation (1) is only valid for monochromatic spectra. GOME measurements however
and most of the reference spectra are convoluted with the instruments slit function
and the assumption of monochromatic spectra breaks down. Furthermore, reference spectra were both shifted by about 0.017 nm towards longer wavelengths,
in order to account for a slight spectral misregistration. Inelastic rotational Raman
scattering (the so-called Ring effect) in the atmosphere acts as a pseudo-absorber in
Equation (2) and this is considered by fitting a corresponding Ring spectrum taken
from Vountas et al. (1998). An instrument correction spectrum is applied to correct
for the slight undersampling of GOME in the UV spectral region (Slijkhuis et al.,
1999).
Slant columns are converted to geometry-independent vertical column amounts
through division by appropriate air mass factors (AMF). Air mass factors describe
the enhanced absorption of a given trace gas due to slant paths of incident light
in the atmosphere. A common definition that is used in our study is given in the
following:
AMF(λ) =
τslant (λ)
τvert (λ)
where
Ino
Itot
(3)
τslant (λ) = ln
∞
σg (λ)C g (z) dz
τvert (λ) =
(4)
(5)
0
where τslant (λ) and τvert (λ) are the slant and vertical optical densities, Itot the total
upwelling intensity including all absorbers, Ino the upwelling intensity except the
trace gas of interest, σg (λ) the cross-section, and C g (z) the layer concentration of
gas g.
The required slant optical densities are derived from radiative transfer calculations (Rozanov et al., 1997) including the pseudo-spherical approximation. Viewing
geometry and geolocation information are taken from GOME data. AMFs depend
further on the reflection properties of the underlying surface (ground or clouds),
the height above sea level (ground or cloud-top), the aerosol scattering and extinction properties, and the atmospheric profiles of temperature, pressure and trace gas
concentrations. A single air mass factor at a reference wavelength is applied, that
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
299
is representative for the DOAS spectral fitting interval. This is discussed in more
detail in Section 4.
The vertical column density (VCD) of a trace species is given by the ratio of the
total slant column amount SC to the total air mass factor AMF. Clouds are treated
in the independent pixel approximation which finally gives:
VCDtotal =
SC + fc · GVC · AMFcloud
(1 − fc ) · AMFground + fc · AMFcloud
(6)
where fc is the cloud coverage between 0 and 1, SC the DOAS-fitted slant column,
GVC the ghost vertical column of a trace species below the cloud-top (derived
from climatological profiles), and AMFgrounds , AMFcloud the air mass factors down
to ground and to cloud-top, respectively.
3. Slant Column Results
3.1.
JULY / AUGUST
2001
The meteorological conditions during the eruptions in 2001 are characterized by
a stable high pressure system over the Southern Mediterranean and relatively constant westerly winds. Cloud coverage was below 10% throughout the entire period,
providing optimum observation conditions.
The most active phase of Mt. Etna began on 20 July 2001 and a first GOME
overpass on 21 July 2001 showed already enhanced levels of SO2 in the east of the
volcano. The next direct overpass of GOME on 24 July 2001 clearly showed an SO2
plume over the Mediterranean (Figure 1, left panel) between Sicily and the Libyan
coast. The maximum slant column content was around 3.6 DU ± 1.05 DU, which
is around a factor of 10 higher than the typical background content. On 25 July,
enhanced levels of SO2 were even found over the Saharan desert in Libya (Figure 1
right panel). The smaller GOME footprint on 25 July 2001 is due to narrow scan
operations on this day. The ground pixel size is then 80 × 40 km2 only.
A second major eruption was observed on 27 July (Figure 2, left panel) where
SO2 column densities of 5.25 DU ± 1.05 DU were detected, while emissions
became lower on 31 July (Figure 2, right panel) and fell back to pre-eruptive
levels after 6 August. The wind direction changed from North-West to West
and enhanced SO2 levels were therefore measured between Sicily and the Greek
coast.
3.2.
OCTOBER / NOVEMBER
2002
The meteorological conditions during the eruptions in fall 2002 were more variable.
Also the overall cloudiness in the Mediterranean area was higher than during the
year before. From 28 October 2002 to 30 October 2002 north-westerly and northern
winds were observed over the sea while easterly directions were dominant over
300
W. THOMAS ET AL.
Figure 1. Slant columns of sulfur dioxide over the Mediterranean Sea as retrieved from GOME
measurements on 24 July 2001 (left panel) and 25 July 2001 (right panel). The smaller GOME
ground track on the right panel is due to narrow scan operations of the GOME instrument
where the ground pixels size is reduced to 80 × 40 km2 . The observed maximum slant column
content on 24 July 2001 was around 3.7 DU (Dobson Units). Backscan data is excluded from
the image.
Figure 2. Slant columns of sulfur dioxide over the Mediterranean Sea as retrieved from GOME
measurements on 27 July 2001 (left panel) and 31 July 2001 (right panel). The observed
maximum slant column content on 27 July 2001 was around 5.4 DU (Dobson Units). Backscan
data is excluded from the image.
the Saharan desert. First GOME measurements of enhanced SO2 columns were
observed on 28 October 2002 (not shown) where GOME was operated in the
static view mode, thus having ground pixels with very limited spatial coverage
of about 1.9 km × 40 km2 only. On 29 October 2002, a direct overpass of GOME
was analyzed and maximum values were found close to the volcano (Figure 3, left
panel). In the following, north-easterly winds transported the bulk of SO2 to the
Algerian coast (Figure 3, right panel). From 31 October 2002 to 1 November 2002
the wind direction changed to West and enhanced levels were found over parts of
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
301
Figure 3. Slant columns of sulfur dioxide over the Mediterranean Sea as retrieved from GOME
measurements on 29 October 2002 (left panel) and 30 October 2002 (right panel). The observed
maximum slant column content on 29 October 2002 was around 12.1 DU (Dobson Units).
Backscan data is excluded from the image.
Figure 4. Slant columns of sulfur dioxide over the Mediterranean Sea as retrieved from GOME
measurements on 31 October 2002 (left panel) and 1 November 2002 (right panel). The observed
maximum slant column content on 1 November 2002 was around 9.3 DU (Dobson Units).
Backscan data is excluded from the image.
Greece and the Aegean sea on the following days (Figure 4). On 2nd November
2002 enhanced levels of SO2 were measured over Greece while another major
eruption was observed on 4 November 2002 (Figure 5). This active phase of Mt.
Etna continued for several months but last heavy eruptions were observed from
space by the end of November 2002.
In summary, the maximum slant columns measured in fall 2002 were by
more than a factor of 2 higher than corresponding maximum values in July
2001 (not taking into account the even higher results from the static view mode
overpass). Table I provides an overview about maximum measured SO2 slant
columns.
W. THOMAS ET AL.
302
Table I. Maximum sulfur dioxide slant column densities as retrieved from GOME
backscatter data
Day
Slant cols. SO2
Wind direction
24 July 2001
25 July 2001
27 July 2001
31 July 2001
28 Oct. 2002
3.6 DU
3.1 DU
5.3 DU
3.2 DU
19.5 DU
±
±
±
±
±
1.05 DU
1.05 DU
1.05 DU
1.0 DU
1.1 DU
North-West
North-West
West
West
North-East
29 Oct. 2002
30 Oct. 2002
31 Oct. 2002
1 Nov. 2002
2 Nov. 2002
4 Nov. 2002
12.2 DU
7.4 DU
5.9 DU
8.8 DU
6.0 DU
13.0 DU
±
±
±
±
±
±
1.16 DU
1.27 DU
1.2 DU
1.2 DU
1.14 DU
1.1 DU
North-West
North
North
West
West
West
Figure 5. Slant columns of sulfur dioxide over the Mediterranean Sea as retrieved from GOME
measurements on 2 November 2002 (left panel) and 4 November 2002 (right panel). The
observed maximum slant column content on 4 November 2002 was around 13 DU (Dobson
Units). Backscan data is excluded from the image.
4. Air Mass Factor Calculation
4.1.
SPECTRAL DEPENDENCY AND IMPACT OF GROUND ALBEDO
Air Mass Factors for SO2 vary across the SO2 fitting window where the strong
ozone absorption in the Huggins bands dominates the spectral structures. Between
315 nm and 327 nm the spectral variation of AMFs is in the order of 15% for a
low ground albedo of 0.05 and moderate sun zenith angles lower than 60◦ . Application of a single AMF for conversion of slant columns into vertical columns
requires therefore that it is a representative value for the given spectral interval.
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
303
Results from a closed-loop test using pure synthetic data indicate that an AMF
calculated around 320 nm is suitable to reproduce the input total content within
1%. The analysis further revealed that this approximation is valid up to solar zenith
angels of about 65◦ . If solar zenith angles become higher, the retrieval from space
suffers from the large path length in stratospheric levels where ozone absorption is
dominant, which considerably reduces the penetration depth of photons in the UV.
As long as the ground albedo remains low, AMFs decrease rapidly to values even
below 0.5 which indicates that GOME is no longer sensitive to SO2 in the lower
troposphere.
Changes of the surface albedo may occur frequently in GOME footprints over
land and ocean and especially at high latitudes (snow/ice versus water). Sulfur
dioxide AMFs are changing by more than a factor of 2 from dark vegetated surfaces
to highly reflecting lower boundaries, which can also be clouds. Here, the surface
albedo is derived from a monthly data base (Koelemeijer et al., 2003) and extracted
values in the UV are between 0.01 (water) and 0.15 (land surfaces). Values are much
higher for snow-covered surfaces but to our knowledge there was no snowfall in
the Mediterranean area in both time periods. The mean values from the data base
are therefore deemed to be representative in the given time frame.
4.2.
AEROSOL LOADING AND SULFUR DIOXIDE PLUME HEIGHT
The calculation of appropriate AMFs for species with major tropospheric loading
(e.g., SO2 , NO2 ) requires knowledge about its vertical distribution. In satellite applications from nadir-looking sensors the actual concentration profile is commonly
unknown and climatological information is widely used. This may, however, cause
significant errors in the calculated column densities and external information about
the trace gas profiles is in favor. Graf et al. (1999) showed that volcanic plumes
can be highly variable in time and space. GOME ground pixels, however, cover a
relatively big area and especially the spatial distribution of volcanic plumes may not
be resolved adequately. Largest uncertainties of retrieved SO2 total column values
are therefore expected close to the volcano. Here, a standard SO2 profile was taken
from Anderson et al. (1986) and SO2 plumes with a maximum content at 2, 3, 4, 5,
and 6 km above sea level were added. The sulfur dioxide AMFs are not sensitive
to the integrated profile content, which was 6 DU for all profiles.
A more difficult problem is the impact of the aerosol loading (optical thickness,
aerosol type) on calculated sulphur dioxide AMFs and column densities. Even
more, the aerosol type and its vertical distribution are typically not independent
from the height of the SO2 plume. Krueger et al. (1995) showed that a neglected
fairly thin aerosol layer may cause a systematic overestimation of the retrieved
total SO2 content. On the other hand, the optical properties of mineral aerosol differ remarkably from that of maritime or urban components (see e.g., Hess et al.
(1998)). The simulation of sulphur dioxide AMFs during and after volcanic outbreaks requires therefore knowledge about the plume height, to a lesser extent its
W. THOMAS ET AL.
304
Table II. Aerosol components and layer definitions used for AMF calculationsa
Maritime
Desert
Urban
Height (km)
Hum.
(%)
Main aerosol
components
Hum.
(%)
Main aerosol
components
Hum.
(%)
Main aerosol
components
0–1
Variab.
Variab.
7–10
10–30
30–100
80
50
50
50
0
0
mar. polluted
mar. polluted
volcanic ash
mar. clean
sulfate
meteoric
50
50
50
50
0
0
desert
desert
volcanic ash
cont. clean
sulfate
meteoric
80
50
50
50
0
0
urban
urban
volcanic ash
cont. clean
sulfate
meteoric
a
The center of the volcanic ash layer is allowed to vary between 2 km and 6 km while
its vertical thickness is held constant at 2 km. Layer boundaries below and above vary
accordingly.
vertical thickness and the aerosol extinction properties. A sensitivity analysis on the
aerosol loading and the plume height is performed using the aerosol models given
in Table. II. Note, that the desert aerosol model applies a lower relative humidity
of 50% in the Planetary Boundary Layer (PBL), reflecting the lower atmospheric
water vapor content of a typical desert scenario.
We first investigate the impact of different aerosol loading scenarios on SO2
air mass factors. The variance of parameters is according to the prevailing conditions during the recent outbreaks of Mt. Etna in summer 2001 and fall 2002 where
SO2 emissions into the troposphere were observed. All calculations were made
for a solar zenith angle of 45◦ (line-of-sight angle and relative azimuth were set
to 0◦ ) and for a surface albedo of 0.05 which is typical in the UV for cloud-free
and snow-free scenes. We simulate the impact of varying aerosol optical densities
(0.3, 0.7) calculated at a reference wavelength of 550 nm. An aerosol extinction
profile taken from Lowtran 7 (Kneizys et al., 1988) was modified in the troposphere in such a way that the aerosol extinction reaches a relative maximum at
the same height levels as given for the corresponding SO2 layer content. A constant background aerosol loading is always assumed above the plume. Mie-type
phase functions were applied in the radiative transfer simulations for all aerosol
modes shown in the following. A SO2 plume with a vertical thickness of 2 km
is assumed. Radiative transfer calculations were performed with a high vertical
resolution of 0.25 km inside the plume (0.5 km below and 1 km elsewhere), in order to model the rapidly changing aerosol optical properties correctly. The aerosol
parametrization follows the aerosol mixing scheme proposed in Hess et al. (1998)
for main components. A new component “volcanic ash” (single scattering albedo =
0.87) was added to the components data base which is based on recommendations
made in McClatchey (1983). It consists of ash particles and sulfuric acid (75%)
droplets.
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
305
Figure 6. Percentage differences between sulfur dioxide AMFs for various aerosol scenarios
and AMFs of an aerosol-free atmosphere as function of plume height. A volcanic ash layer
and different aerosol models below the volcanic ash layer with a total aerosol optical thickness
of 0.3 and 0.7 are simulated. The center height of the volcanic ash layer varies similar to the
corresponding SO2 plume height.
In addition to varying aerosol optical properties the height of the maximum
aerosol extinction changes from 2 to 6 km above sea level. Three different scenarios were analyzed: AMF calculations over water surfaces using maritime aerosol
components, while a desert aerosol model and a continental polluted aerosol model
were used to represent different but typical atmospheric conditions in the larger
environment of Mt. Etna.
Figure 6 shows the percentage change of sulfur dioxide AMFs (with respect to
aerosol-free conditions) under the presence of varying aerosol loading scenarios as
a function of plume height. For a moderate aerosol optical thickness of 0.3 changes
with the aerosol type are in the order of ±5%. Both the aerosol-free simulations
and the simulations with aerosol were performed for the same plume heights and no
significant additional variation with plume height was observed. Note however that
the presence of desert aerosol in the PBL has the opposite impact (lower AMFs)
than the presence of maritime or urban aerosols (higher AMFs). A higher aerosol
loading (aerosol optical thickness 0.7) causes generally lower sulfur dioxide AMFs
and changes up to −20% occur for the presence of desert aerosols in the lower
troposphere. Relative changes of AMFs with plume height however remain small.
On the other hand relatively large changes of absolute sulfur dioxide AMFs with
plume height were observed (Figure 7). For low aerosol loading an almost linear
increase by a factor of 2 is observed if the plume maximum increases from 2 km to
6 km above ground level. Sulfur dioxide AMFs behave similar for higher aerosol
306
W. THOMAS ET AL.
Figure 7. Variation of sulfur dioxide AMFs with aerosol loading as function of plume height.
A volcanic ash layer and different aerosol loading scenarios below the volcanic ash layer with
total aerosol optical thicknesses of 0.3 and 0.7 are simulated. The center height of the volcanic
ash layer varies similar to the corresponding SO2 plume height.
loading, although absolute values are lower, while lowest AMFs are observed for
the desert aerosol scenario (highest absorption). An improper knowledge of plume
height will therefore cause large uncertainties in retrieved total SO2 column values.
The impact of the aerosol type and the aerosol optical thickness is lower in general
but may be in the same order if a moderate maritime aerosol loading is assumed
while a strong desert aerosol loading is present.
A further analysis of the impact of the aerosol composition of the volcanic
ash layer showed that no significant impact can be observed for lower aerosol
optical thickness. In case of higher aerosol optical thickness (0.7) an additional
water-soluble aerosol, sulphate aerosol (although not representative for sulfuric
droplets in the troposphere outside the Arctics) or soot aerosol component in the
plume does not cause significant changes of the results presented above. Here, we
added 10% and 50% (in number mixing ratios) other aerosol components to the ash
layer. If however the more absorbing dust-like water-insoluble aerosol component
(single scattering albedo = 0.67) is added, the impact of the composition of the
ash layer becomes higher. We assumed a contribution of 1% and 10% (number
mixing ratios) and SO2 AMFs decreased in the order of 3–4% but between 30%
(low plume height) and 18% (high plume height), respectively, if compared against
simulations with the aerosol component “volcanic ash” only. This further translates
into an underestimation of sulphur dioxide total columns of the same range. We
therefore conclude that the aerosol composition of the ash layer is only important
if the aerosol optical thickness is remarkably higher than 0.3 and large particles are
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
307
present. This will be the case close to the volcano where the SO2 retrieval is crucial
due to the GOME pixel size. Note, that the dust-like aerosol consists of comparably
large particles which will be removed faster by dry and wet deposition than smaller
aerosol particles. Such information about the particle size can be estimated from
e.g. Lidar measurements but will rapidly change with the distance from the volcano.
Moreover, the maximum SO2 layer content and the maximum aerosol extinction
coefficient may not be observed at same height levels. Large aerosol particles can be
removed more quickly than gaseous emissions from higher atmospheric levels by
dry deposition, and this is further analyzed. We used the basic scenarios presented in
Table II but the aerosol plume now resides 1km below the SO2 peak concentration.
AMF results however are comparable (in both its absolute and relative changes)
to those retrieved for the cases discussed before. It is therefore concluded that the
main impact on AMFs is with the height of the SO2 plume while an uncertainty of
the height level of the maximum aerosol extinction is typically of less importance.
5. Total Column Results and Validation
5.1.
TRAJECTORY ANALYSIS
To overcome the problem of the unknown plume height, we investigated meteorological analysis data and data from other space-borne sensors, that can provide
information, both about cloud type and cloud-top height. During both the outbreaks
of Mt. Etna its smoke and ash plume was evident in the visible and infrared channels of AVHRR (NOAA-14, NOAA-16), as mid-level cloud between 700 hPa and
400 hPa above sea level. Cloud levels were derived from the AVHRR processing
scheme over land and ocean (APOLLO) which was recently updated by Kriebel
et al. (2003). However, the sensitivity analysis presented in Section 4.2 gave evidence for a more precise knowledge about the plume height.
Forward and backward trajectories were calculated using the FLEXTRA model
described in Stohl et al. (1999). It is driven by analyses of the European Center
for Medium Range Weather Forecasts. Ensembles of 3D forward trajectories were
released at Mt.Etna at different pressure levels from 700 hPa to 200 hPa. By matching the parcel trajectories with the first guess GOME SO2 retrievals the height of
the SO2 plume is determined. The effect of the uncertainty of the plume height on
the retrieval is discussed in Section 4.2. Additionally, ensembles of backward trajectories were calculated from GOME observations with elevated SO2 levels. This
allows to proof the volcanic origin of the SO2 , to determine the effective emission
height over Mt. Etna and to get insight into the spatial and temporal evolution of
the SO2 plume. Table III gives an overview about days and GOME overpasses that
were investigated in detail.
Forward runs started on 24 July 2001 at 1020 UTC at Mt. Etna in several
height levels between 200 hPa and 700 hPa. Figure 8 clearly shows that SO2 was
transported within 24 h to South-East directions between 500 hPa and 700 hPa.
W. THOMAS ET AL.
308
Table III. Days, GOME overpasses and geographical position of sulfur dioxide
for which trajectory analysis was performed
Day
GOME orbit number
Traj. Modelling
Location
24 July 2001
25 July 2001
31 Oct. 2002
1 Nov. 2002
1 Nov. 2002
2 Nov. 2002
32727
32741
39368
39382
39382
39396
Forward
Backward
Forward
Backward
Forward
Backward
37.70◦ N, 15.00◦ E
29.02◦ N, 20.51◦ E
37.70◦ N, 15.00◦ E
38.51◦ N, 23.29◦ E
37.70◦ N, 15.00◦ E
39.80◦ N, 24.52◦ E
Figure 8. Forward trajectory analysis started on 24 July 2001 (1020 UTC) at Mt. Etna (upper
panel) and corresponding height levels as a function of time (lower panel). The forward trajectories released at 500 hPa and 600 hPa are crossing the region with enhanced sulfur dioxide
levels over Libya (see Figure 1) 24 h later in height levels between 620 hPa and 750 hPa.
Trajectories stop in the lower panel if the given geographic area is left.
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
309
Figure 9. Backward trajectory analysis started on 25 July 2001 (1020 UTC) at Libya (upper
panel) and corresponding height levels as a function of time (lower panel). The backward
trajectory released at 700 hPa is crossing the region around Mt. Etna at 600 hPa which is
consistent with results from the forward trajectory modelling. Trajectories stop in the lower
panel if the given geographic area is left.
Enhanced levels of SO2 at the Libyan coast can therefore be attributed to emissions
from Mt. Etna on 23 July 2001 and 24 July 2001. The corresponding backward
trajectories where released on 25 July 2001 (Figure 9) at same height levels at the
SO2 spot shown in Figure 1 (for details see Table III). These results are consistent
with results from the forward modelling and support the transport of volcanic
emissions between 500 hPa and 600 hPa from the Etna to the given location. The
height of the SO2 plume that is needed for reliable AMF computations was then
set to 4 km above sea level which is a representative value for this period.
Results for the second event in Fall 2002 are shown in Figures 10 and 11 where
enhanced levels of SO2 were observed over Greece (see Figure 4, right panel).
310
W. THOMAS ET AL.
Figure 10. Forward trajectory analysis started on 31 October 2002 (1020 UTC) at Mt. Etna
(upper panel) and corresponding height levels as a function of time (lower panel). The forward
trajectories released between 500 hPa and 600 hPa are crossing the region around Athens
(Greece) (see also Figure 4) within 18 h. Height levels of the two upper trajectories decrease
by about 50 hPa in this time frame. Trajectories stop in the lower panel if the given geographic
area is left.
Forward runs on 31 October 2002 again started at 1020 UTC at Mt. Etna in a number of height levels. The analysis revealed that strong westerly winds transported
the emissions of Etna between 600 hPa and 400 hPa within 24 h to regions easterly of Greece. The trajectory released at 500 hPa follows a loop over the Eastern
Mediterranean which may occur if wind directions are changing in the area within
the time interval of the simulation. The corresponding results from the backward
trajectory analysis from 1 November 2002 (see Figure 11) confirm again the results from the forward modelling. In general, volcanic emissions in fall 2002 were
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
311
Figure 11. Backward trajectory analysis started on 1 November 2002 (1020 UTC) at Greece
near Athens (upper panel) and corresponding height levels as a function of time (lower panel).
The backward trajectory released between 300 hPa and 500 hPa are crossing the region around
Mt. Etna within 18 h. Height levels of the two upper trajectories decrease by about 50 hPa in
this time frame, which is consistent with results from the forward modelling. Trajectories stop
in the lower panel if the given geographic area is left.
emitted into higher atmospheric levels and observed SO2 levels were higher, which
underlines that these eruptions were more powerful than in summer 2001.
5.2. GOME
MEASUREMENTS AND ERROR BUDGET
Sulfur dioxide AMFs and vertical columns were calculated taking into account the
information from the trajectory analysis (SO2 plume height) and the lidar measurements (fall 2002 only). The summer scenarios were calculated using an average
W. THOMAS ET AL.
312
single SO2 profile with a peak at 4 km above sea level while the fall scenario has a
relative maximum of SO2 at 5 km above sea level. Corresponding aerosol extinction
profiles were set accordingly. The most plausible maritime aerosol model (aerosol
optical thickness 0.3) was applied in the PBL for AMF calculations. As discussed
in Section 4.2, this model can be applied also for an urban aerosol loading but will
lead to a systematic underestimation of sulphur dioxide columns by about 10% for
the same aerosol optical thickness over desert surfaces. On average, the impact of
improperly known aerosol properties on calculated trace gas total columns is in the
order of <10%, provided that the aerosol optical thickness remains below 0.3.
In the area of interest AMFs vary smoothly with the solar zenith angle. For the
summer scenario SO2 AMFs around the plume are between 1.4 an 1.6 while AMFs
increase in fall 2002 to values around 1.7 to 2.0. As discussed in Section 4.2 AMFs
my even be lower close to the volcano, due to the presence of large particles in
the plume. Application of simple geometric AMFs (2.3–2.7) however will systematically underestimate total columns. Maximum SO2 total columns for the orbits
shown in Figures 1–5 are summarized in Table IV. Given error numbers are mainly
due to the error of the spectral fitting (rms-error) while the uncertainty due to the
aerosol loading is not covered (bias).
Application of trajectory analysis technique may considerably improve the
knowledge of plume height but the remaining uncertainty will typically be in the
order of ±1 km. If the uncertainty of the SO2 plume height is assumed to be in
this range, the corresponding uncertainty of AMFs over water surfaces (maritime
aerosol model with aer. opt. density 0.3 assumed) is about ±0.2 or around ±15%
(summer 2001) and ±12% (fall 2002). Relative changes of AMFs with changing
plume height are similar for the higher aerosol optical thickness of 0.7 and other
aerosol models (except for the desert model with high optical thickness where
relative changes of AMFs with plume height are only in the order of 0.15). This
Table IV. Maximum sulfur dioxide vertical column
densities as retrieved from GOME backscatter data
Day
VCD SO2
24 July 2001
25 July 2001
27 July 2001
31 July 2001
29 Oct. 2002
30 Oct. 2002
31 Oct. 2002
1 Nov. 2002
2 Nov. 2002
4 Nov. 2002
2.2 DU ± 0.64 DU
1.9 DU ± 0.65 DU
3.7 DU ± 0.77 DU
2.0 DU ± 0.65 DU
6.0 DU ± 0.54 DU
3.7 DU ± 0.63 DU
3.3 DU ± 0.73 DU
4.3 DU ± 0.69 DU
3.0 DU ± 0.57 DU
4.5 DU ± 0.58 DU
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
313
additional error is not covered by error levels given above but needs to be considered, in order to draw a more realistic image of the accuracy of retrieved sulphur
dioxide levels. Total error levels without taking into account the uncertainty of
the plume height are not better than ±30% for the first episode in Summer 2001
while errors are in the order of ±20% for the second episode in Fall 2002. A possible bias of at least 12% needs to be taken into account for insufficiently known
profile data, mainly the plume height. A further uncertainty of about ± 30% must
to be assumed for AMFs of pixels close to the volcano, due to the insufficiently
known aerosol composition of the volcanic ash layer (see Section 4.2).
5.3.
COMPARISON OF SPACE - BORNE DATA AND GROUND - BASED RESULTS
Ground-based measurements of sulphur dioxide (Brewer spectrophotometer) and
an aerosol extinction profile taken from Lidar measurements of a Raman Lidar at
Thessaloniki (Greece) are partially available for the second active phase of Mt. Etna
in fall 2002 (Balis et al., 2003).
We analyzed nighttime lidar measurements (Raman-Lidar) at Thessaloniki from
31 October 2002 to 1 November 2002 (see Figure 12). Here, an increasing aerosol
extinction coefficient between 4500 m and 5500 m above sea level is attributed to
aerosol particles of volcanic origin. In the same figure the extinction-to-backscatter
ratio is shown, a quantity usually called as lidar ratio, which characterizes the
Figure 12. Nighttime lidar measurements from Thessaloniki station on 31 October 2002 showing the backscatter coefficient (left panel) at wavelengths 355 nm and 532 nm, the aerosol
extinction coefficient at 355 nm (middle panel), and the lidar ratio at 355 nm (right panel). The
enhancement of the extinction coefficient between 4500 m and 5500 m is attributed to volcanic
ash emitted by Mt. Etna.
314
W. THOMAS ET AL.
micro-physical properties of aerosols (particle size, refractive index, shape). As
it is evident in Figure 12 the lidar ratio increases between 4500 m and 5500 m
above sea level from 10 sr to about 40 sr indicating the presence of more absorbing
aerosols of volcanic origin in that layer relative to layers below. However, the early
arrival of aerosols was not due to direct transport from Mt. Etna to Greece but due
to the transport of aged eruptive material from the beginning of the active period
on 28 October 2002. The aerosol was first transported to the South-West but a
change of the prevailing wind direction to west transported the aerosol plume back
Figure 13. Forward trajectory analysis started on 1 November 2002 (1020 UTC) at Mt. Etna
(upper panel) and corresponding height levels as a function of time (lower panel). The forward
trajectories released at 400 hPa and 500 hPa are crossing the region around Thessaloniki
(Greece) indicated by the red symbol (see also Figure 5). Height levels are slightly decreasing
by less than 20 hPa within 24 h. Trajectories stop in the lower panel if the given geographic
area is left.
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
315
to east while passing again the Etna in the North of Sicily. It is therefore unlikely
that large particles were still present in the plume, and this is again confirmed by
the Lidar measurements. The stronger backscatter at 355 nm compared to 532 nm
in the layer between 4500 m and 5500 m indicates an Angstrøm exponent larger
than one which is typical for small particles (see Figure 12). Consequently, the
SO2 retrieval was performed without a dust-like aerosol component in the volcanic
plume.
Figure 14. Backward trajectory analysis started on 2 November 2002 (1020 UTC) near Thessaloniki (Greece) indicated by the red symbol (upper panel) and corresponding height levels
as a function of time (lower panel). The backward trajectory released between 300 hPa and
500 hPa are crossing the region around Mt. Etna within 24 h without changing height levels
significantly, which is consistent with results from the forward modelling. Trajectories stop in
the lower panel if the given geographic area is left.
316
W. THOMAS ET AL.
GOME measurements indicate the presence of SO2 over Thessaloniki on 2nd
November 2002 (see Figures 4 and 5). Forward trajectories were released from
Mt. Etna between 300 hPa and 700 hPa on 1 November 2002 at 1020 UTC and the
region of Thessaloniki was reached by the 400 hPa and 500 hPa trajectories 24 h later
(Figure 13). The results from the backward runs starting on 2nd November 2002 at
same height levels at 1020 UTC are consistent with results from the forward runs
and confirm the origin of air masses from the region around Mt. Etna (Figure 14).
Zerefos et al. (2000) already showed the important role of Mt. Etna for the SO2
budget above Northern Greece.
Ground-based measurements of sulphur dioxide are available from Brewer spectrophotometer measurements. The Brewer instruments are however designed to
measure primarily the atmospheric ozone content while the retrieval of sulphur
dioxide is a side effect of these measurements, which is necessary for correct
ozone measurements. A larger uncertainty of sulphur dioxide amounts retrieved
from Brewer data is therefore expected, as it is discussed in detail by Fioletov
et al. (1998). Here, the Brewer measurements indicate the arrival of the sulphur
dioxide plume on 01 November 2002 and the maximum total content of about
5–7 DU is reached on 02 November 2002 (see Figure 15). The GOME-measured
SO2 total content above Thessaloniki was about 3 DU on 2nd November 2002
(Figure 16) which is roughly half of the results from the ground-based measurements. However, GOME pixels cover a much larger area (here: large parts over
Figure 15. Brewer measurements of sulfur dioxide from Thessaloniki station between 31
October 2002 and 2 November 2002. Vertical arrows indicate the GOME observation time.
The increase of sulfur dioxide levels from 2–4 DU on 31 October 2002 to 5–7 DU on 2
November 2002 is consistent with GOME total column measurements of about 3 DU (see text
and also Figure 10).
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
317
Figure 16. Vertical columns of sulfur dioxide over the Mediterranean Sea as retrieved from
GOME measurements on 2 November 2002. The observed maximum content was around 3
DU (Dobson Units). The presence of enhanced levels of sulfur dioxide at the African coast line
and over the Eastern Mediterranean is in line with forward trajectory runs from 31 October
2002 (see Figure 10). Backscan data is excluded from the image.
the Agean Sea) and GOME is especially not sensitive to the SO2 in the lower
atmosphere under low albedo conditions. Sulphur dioxide AMFs are then typically in the order of 0.2 indicating that GOME is an order of magnitude less
sensitive to SO2 close to the surface. The Brewer measurements indicate an SO2
background content of about 2–4 DU before the arrival of the volcanic plume.
The additional SO2 content due to transport from Mt. Etna is therefore in the order of 3 DU which is in good quantitative agreement with corresponding GOME
results.
6. Conclusion
We analyzed GOME backscatter measurements during two eruptive periods of
Mt. Etna for the atmospheric SO2 content. The main problem of space-borne SO2
318
W. THOMAS ET AL.
measurements is the retrieval of its columnar content. We could show that the large
uncertainty of SO2 total column measurements during and after volcanic eruptions
can be improved if information about the SO2 profile and especially the plume
height of volcanic emissions is available from other sources. A knowledge of the
plume height not better than 1 km causes an uncertainty of derived SO2 values in
the order of 15%. Also the aerosol loading is of some importance if the aerosol
optical thickness becomes larger than 0.3. The relatively large GOME footprints
however make it unlikely that such values occur frequently. The knowledge of the
aerosol loading in the PBL is more important if desert aerosol components are
likely to be present. If a desert aerosol component with an aerosol optical thickness
of 0.3 is neglected the retrieval will underestimate the SO2 total column by less
than about 5–10%. An overestimation of 5% is however expected for maritime and
urban aerosol components of the same optical thickness.
GOME results from the eruptive period of Mt. Etna in fall 2002 were compared against ground-based data from the Brewer station at Thessaloniki (Greece).
Forward and backward trajectory analysis data gave evidence for the presence of
volcanic emissions from Mt. Etna in the region of Thessaloniki. As expected, the
SO2 total columns measured from ground are higher than corresponding GOME
measurements but the relative increase of SO2 levels due to the volcanic emissions
was in the same order as GOME-derived SO2 total column results.
For both eruptive periods of Mt. Etna the trajectory analysis gave evidence that
there was no injection of SO2 into stratospheric levels. Therefore, any climatological
effect of these eruptions is not expected. The estimation of total volcanic SO2
emissions from GOME backscatter data remains difficult because the overpass
cycle of GOME is about three days which is in the same order as the tropospheric
lifetime of SO2 . Furthermore, estimation of total emitted SO2 mass suffers from
inadequate coverage of the satellite data in time and space. Measurements of the
global volcanic SO2 budget with space-borne instruments like GOME are therefore
limited, but re-analysis of historical data can be performed by using information
from external sources and error margins may decrease remarkably. Instruments like
the upcoming GOME-2/METOP (launch planned in late 2005) will improve the
situation through a better spatial and temporal coverage, but the need for auxiliary
data remains until more advanced retrieval methods for e.g., the SO2 profile and
the aerosol loading from these sensors are available.
Acknowledgements
This work was partially supported within the framework of the EUMETSAT Satellite Application Facility on Ozone Monitoring. The lidar measurements at Thessaloniki were performed in the frame of the EARLINET project funded by the
European Union. The assistance of V. Amiridis is greatly acknowledged. Special
thanks to S. Slijkhuis (DLR) for providing an undersampling correction spectrum.
A. Stohl kindly provided the FLEXTRA model and we thank the ECMWF for
ON THE RETRIEVAL OF VOLCANIC SULFUR DIOXIDE EMISSIONS
319
analyses data. Helpful discussions with M. Eisinger (ESA-ESTEC) on sulfur dioxide retrieval are gratefully acknowledged.
References
Aliwell, S., Roozendael, M. V., Johnston, P., Richter, A., Wagner, T., Arlander, B., Burrows, J., Fish,
D., Jones, R., Lambert, J.-C., Pundt, I., and Tornkvist, K., 2002: Analysis for BrO in zenithsky spectra: An intercomparison exercise for analysis improvement, J. Geophys. Res. 107, D14,
10.1029/2001JG000329.
Anderson, G., Clough, S., Kneizys, F., Whetwynd, J., and Shettle, E. P., 1986: AFGL Atmospheric
Constituents Profiles (0–120 km). Report AFGL-TR-86-1001 AD175173, Air Force Geophysics
Laboratory, Hanscom, MA., U.S.A.
Balis, D., Amiridis, V., Zerefos, C., Gerasopoulos, E., Andreae, M., Zanis, P., Kazantzidis, A.,
Kazadzis, S., and Papayannis, A., 2003: Raman lidar and sunphotometric measurements of aerosol
optical properties over Thessaloniki, Greece during a biomass burning episode, Atm. Env. 37,
4529–4538.
Burrows, J., Dehn, A., Deters, B., Himmelmann, S., Richter, A., Voigt, S., and Orphal, J., 1998: Atmospheric remote-sensing reference data from GOME: Part 1. Temperature-dependent Absorption
cross-sections of NO2 in the 231–794 nm range, J. Quant. Spec. Radiat. Trans. 60, 1025–1031.
Burrows, J., Richter, A., Dehn, A., Deters, B., Himmelmann, S., Voigt, S., and Orphal, J., 1999a: Atmospheric Remote-Sensing Reference Data from GOME: Part 2. Temperature-dependent Absorption
cross-sections of O3 in the 231–794 nm range, J. Quant. Spec. Radiat. Trans. 61, 509–517.
Burrows, J., Weber, M., Buchwitz, M., Rozanov, V., Ladstätter-Weißenmayer, A., Richter, A., de
Beek, R., Hoogen, R., Bramstedt, K., Eichmann, K.-U., Eisinger, M., and Perner, D., 1999b: The
Global Ozone Monitoring Experiment (GOME): Mission concept and first scientific results, J.
Atm. Sci. 56, 151–175.
Eisinger, M. and Burrows, J., 1998: Tropospheric sulfur dioxide observed by the ERS-2 GOME
instrument, Geophys. Res. Lett. 25, 4177–4180.
Fioletov, V., Griffioen, E., Kerr, J., Wardle, D., and Uchino, O., 1998: Influence of volcanic sulfur
dioxide on spectral UV irradiance as measured by Brewer spectrophotometers, Geophys. Res.
Lett. 25, 1665–1668.
Graf, H.-F., Feichter, J., and Langmann, B., 1997: Volcanic sulfur emissions: Estimates of source
strength and its contribution to the global sulfate distribution, J. Geophys. Res. 102, 10727–10738.
Graf, H.-F., Herzog, M., Oberhuber, J., and Textor, C., 1999: Effect of environmental conditions on
volcanic plume rise, J. Geophys. Res. 104, 24309–24320.
Hansen, J., Laci, A., Ruedy, R., and Sato, M., 1992, Potential climate impact of Mt. Pinatubo eruption,
Geophys. Res. Lett. 19, 215–218.
Hess, M., Köpke, P., and Schult, I., 1998: Optical properties of aerosols and clouds: The Software
Package OPAC, Bull. Am. Meteorol. Soc. 79, 831–844.
Kneizys, F., Shettle, E., Abreu, L., Chetwynd, J., Anderson, G., Gallery, W., Selby, J., and Clough, S.,
1988: ‘Users Guide to LOWTRAN 7’. Air Force Geophysics Laboratory, Environmental Research
Papers, no.1010, afgl-tr-88-0177 edition.
Koelemeijer, R., de Haan, J. F., and Stammes, P., 2003: A database of spectral surface reflectivity
in the range 335–772 nm derived from 5.5 years of GOME observations, J. Geophys. Res. 108,
4070, doi:10.1029/2002JD0024.
Koelemeijer, R. and Stammes, P., 1999: Validation of Global ozone monitoring experiment cloud
fractions for accurate ozone column retrieval, J. Geophys. Res. 104, 18801–18814.
Kriebel, K., Gesell, G., Kästner, M., and Mannstein, H., 2003: The cloud analysis tool APOLLO:
Improvements and validation, Int. J. Rem. Sens. 24, 2389–2408.
320
W. THOMAS ET AL.
Krueger, A., 1983: Sighting of El Chichon sulfur dioxide clouds with the Nimbus 7 total ozone
mapping spectrometer, Science 220, 1377–1379.
Krueger, A., Walter, L., Bhartia, P., Schnetzler, C., Krotkov, N., Sprod, I., and Bluth, G., 1995: Volcanic
sulfur dioxide measurements from the total ozone mapping spectrometer instruments, J. Geophys.
Res. 100, 14057–14076.
Kuze, A. and Chance, K., 1994: Analysis of cloud-top height and cloud coverage from satellites using
the O2 A and B Bands, J. Geophys. Res. 99, 14481–14491.
McClatchey, R., 1983: ‘Report of the experts meeting on aerosol and their climatic effects’. Technical
Report WCP-55, World Meteorological Organization, Boulder, Colorado.
McPeters, R., 1993: The atmospheric SO2 budget for Pinatubo derived from NOAA-11 SBUV/2
spectral data, Geophys. Res. Lett. 18, 1971–1974.
Rozanov, V., Diebel, D., Spurr, R., and Burrows, J. P., 1997: GOMETRAN: Radiative transfer model
for the satellite project GOME, the plane-parallel version, J. Geophys. Res. 102, 16683–16695.
Slijkhuis, S., Bargen, A. v., Thomas, W., and Chance, K., 1999: ‘Calculation of undersampling correction spectra for DOAS spectral fitting’, in ESAMS’99 – European Symposium on Atmospheric
Measurements from Space, edited by ESA, Earth Science Division, Noordwijk, The Netherlands,
WPP-161, pp. 563–569.
Stohl, A., Haimberger, L., Scheele, M., and Wernli, H., 1999: An intercomparison of results from
three trajectory models, Meteorol. Applications 8, 127–135.
Thomas, W., Baier, F., Erbertseder, T., and Kästner, M., 2003: Analysis of the Algerian severe weather
event in November 2001 and its impact on ozone and nitrogen dioxide distributions, Tellus B 55,
993–1006.
Vountas, M., Rozanov, V., and Burrows, J., 1998: Ring effect: impact of rotational Raman scattering
on radiative transfer in Earth’s Atmosphere, J. Quant. Spec. Radiat. Trans. 60, 943–961.
Zerefos, C., Ganev, K., Kourtidis, K., Tzortziou, M., Vasaras, A., and Syrakov, E., 2000: On the origin
of SO2 above Northern Greece, Geophys. Res. Lett. 28, 365–368.