SAOZ measurements of NO2 at Aberystwyth.
ABSTRACT We present in this paper fifteen years' measurements, from March 1991 to September 2005, of stratospheric NO2 vertical columns measured by a SAOZ zenith-sky visible spectrometer. The instrument spent most of its time at Aberystwyth, Wales, with occasional excursions to other locations. The data have been analysed with the WinDOAS analysis program with low-temperature high-resolution NO2 cross-sections and fitting a slit function to each spectrum. Because of a change in detector in May 1998 there is some uncertainty about the relative changes before and after this date, which are partially constrained by the results of an intercomparison exercise. However, the effect of the Mt Pinatubo aerosol cloud is very evident in the data from 1991-94, with a decrease of 10% in NO2 in the summer of 1992 (the SAOZ was located in Lerwick, Scotland during the winter of 1991-92 and observed very low NO2 values but these cannot be directly compared to the Aberystwyth data). To focus more on interannual and long-term variations in NO2, a seasonal variation comprising an annual and semi-annual component was fitted to the morning and evening twilight separately from 1995 to the present. This fit yielded average NO2 columns of 4.08 x 10(15) cm(-2) and 2.68 x 10(15) cm(-2) for the evening and morning twilight, respectively, with a corresponding annual amplitude of +/-2.08 x 10(15) cm(-2) and +/-1.50 x 10(15) cm(-2). Departures from the fitted curve show a trend of 6% per decade, consistent with that reported elsewhere, for the period 1998-2003, but in the past two years a distinct interannual variation of amplitude of approximately 8% has emerged.
- [show abstract] [hide abstract]
ABSTRACT: The NO(2) abundance in the stratosphere has been determined from ground-based spectra of the rising and setting sun and moon and of the twilight sky near 4500 angstroms. The spectra were taken at the Fritz Peak Observatory, at an altitude of 3 kilometers in the Colorado mountains. Separation of the stratospheric contribution requires observations at a relatively unpolluted site; direct measurement of the tropospheric absorption in the Colorado mountains often yields an upper limit on the tropospheric mixing ratio of 0.1 part per billion. The stratospheric NO(2) abundance is two to three times greater at night than during the day and increases significantly during the course of a sunlit day; these changes are related to photolytic decomposition of NO(2) and N(2)O(5) in the daytime stratosphere. Absorption by NO(3) was sought but not found; the results set an upper limit of 2 percent on the nighttime abundance ratio of NO(3) to NO(2) in the stratosphere.Science 09/1975; 189(4202):547-9. · 31.03 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: 1] We describe the retrieval of stratospheric NO 2 vertical column densities from the Global Ozone Monitoring Experiment (GOME) aboard the ERS-2 satellite. Different differential optical absorption spectroscopy (DOAS) evaluations are compared in order to investigate uncertainties caused by the diffuser plate. An improved version of our algorithm to separate the tropospheric and stratospheric fraction of NO 2 from GOME satellite data is described and is used to extract a long term data set of stratospheric NO 2 (1996–2000). In addition, the average seasonal variation in the global distribution is determined, which allows us to monitor and investigate specific aspects of stratospheric chemistry, in particular the interhemispheric comparison of stratospheric NO 2 . In contrast to other satellite observations (e.g., SAGE II, OSIRIS), GOME observations of stratospheric NO 2 include the lower stratosphere. In general, our observations are in agreement with previous measurements and confirm the current knowledge of stratospheric nitrogen chemistry.J. Geophys. Res. 01/2004; 109.
- [show abstract] [hide abstract]
ABSTRACT: We present an analysis of the impact of heterogeneous chemistry on the partitioning of nitrogen species measured by the Upper Atmosphere Research Satellite (UARS) instruments. The UARS measurements utilized include N2O, HNO3, and ClONO2 from the cryogenic limb array etalon spectrometer (CLAES), version 7, and temperature, methane, ozone, H2O, HCl, NO and NO2 from the halogen occultation experiment (HALOE), version 18. The analysis is carried out for the UARS data obtained between January 1992 and September 1994 in the 100 - to 1- mbar (approx. 17-47 km) altitude range and over 10 deg. latitude bins from 70 deg S to 70 deg N. The spatiotemporal evolution of aerosol surface area density (SAD) is adopted from analysis of the Stratospheric Aerosol and Gas Experiment (SAGE) II data. A diurnal steady state photochemical box model, constrained by the temperature, ozone, H2O, CH4, aerosol SAD, and columns Of O2 and O3 above the point of interest, has been used as the main tool to analyze these data. Total inorganic nitrogen (NO(y)) is obtained by three different methods: (1) as a sum of the UARS-measured NO, NO2, HNO3, and CIONO2; (2) from the N2O-NO(y) correlation, and (3) from the CH4-NO(y) correlation. To validate our current understanding of stratospheric heterogeneous chemistry for post-Pinatubo conditions, the model-calculated monthly averaged NO(x)/NO(y) ratios and the NO, NO2, and HNO3 profiles are compared with the UARS-derived data. In general, the UARS-constrained box model captures the main features of nitrogen species partitioning in the post-Pinatubo years, such as recovery of NO(x) after the eruption, their seasonal variability and vertical profiles. However, the model underestimates the NO2 content, particularly in the 30- to 7-mbar (approx. 23-32 km) range. Comparisons of the calculated temporal behavior of the partial columns of NO2 and HNO3 and ground-based measurements at 45 deg S and 45 deg N are also presented. Our analysis indicates that ground-based and HALOE v.18 measurements of the NO2 vertical columns are consistent within the range of their uncertainties and are systematically higher (up to 50%) than the model results at midlatitudes in both hemispheres. Reasonable agreement is obtained for HNO3 columns at 45 deg S, suggesting some problems with nitrogen species partitioning in the model. Outstanding uncertainties are discussed.05/1999;
SAOZ measurements of NO2at Aberystwyth
G. Vaughan,*aP. T. Quinn,aA. C. Green,abJ. Bean,aH. K. Roscoe,c
M. van Roozendaeldand F. Goutaile
Received 11th August 2005, Accepted 19th December 2005
First published as an Advance Article on the web 18th January 2006
We present in this paper fifteen years’ measurements, from March 1991 to September 2005, of
stratospheric NO2vertical columns measured by a SAOZ zenith-sky visible spectrometer. The
instrument spent most of its time at Aberystwyth, Wales, with occasional excursions to other
locations. The data have been analysed with the WinDOAS analysis program with low-
temperature high-resolution NO2cross-sections and fitting a slit function to each spectrum.
Because of a change in detector in May 1998 there is some uncertainty about the relative changes
before and after this date, which are partially constrained by the results of an intercomparison
exercise. However, the effect of the Mt Pinatubo aerosol cloud is very evident in the data from
1991–94, with a decrease of 10% in NO2in the summer of 1992 (the SAOZ was located in
Lerwick, Scotland during the winter of 1991–92 and observed very low NO2values but these
cannot be directly compared to the Aberystwyth data). To focus more on interannual and long-
term variations in NO2, a seasonal variation comprising an annual and semi-annual component
was fitted to the morning and evening twilight separately from 1995 to the present. This fit
yielded average NO2columns of 4.08 ? 1015cm?2and 2.68 ? 1015cm?2for the evening and
morning twilight, respectively, with a corresponding annual amplitude of ?2.08 ? 1015cm?2and
?1.50 ? 1015cm?2. Departures from the fitted curve show a trend of 6% per decade, consistent
with that reported elsewhere, for the period 1998–2003, but in the past two years a distinct
interannual variation of amplitude of B8% has emerged.
Nitrogen dioxide (NO2) is an important trace gas in the
stratosphere due to its significant role in the photochemistry
of ozone. Stratospheric ozone plays a critical role in the
atmosphere by absorbing most of the biologically damaging
ultraviolet radiation from the sun before it reaches the Earth’s
surface. Between around 25 km and 40 km altitude NO2is
involved in a catalytic cycle that accounts for almost half the
ozone removed by gas-phase reactions.1Below 25 km NO2
moderates ozone loss caused by active chlorine and hydrogen
by converting them into their inactive reservoir forms. Thus,
monitoring of stratospheric NO2is an important adjunct to
international efforts to monitor the health of the ozone layer.
One of the longest duration time series of stratospheric NO2
measurements is that from Lauder, New Zealand, which
extends from 1980 to the present. Analysis of this record
shows that the atmospheric column of NO2is increasing at
approximately 5% per decade at 451S.2Measurements from
similar latitudes in the Northern Hemisphere (using infrared
spectroscopy) show a similar trend.3,4This is twice the rate of
increase of nitrous oxide (N2O), its main source gas, and could
be an indication of changes to the overall atmospheric circula-
tion.5Using a combination of photochemical and 3-D chem-
istry transport models, McLinden et al.6were able to show
that the Lauder trend can be understood if the increase in N2O
is combined with the decrease in ozone, the change in odd
nitrogen partitioning due to increased chlorine concentrations,
and variations in volcanic aerosols. Continued measurements
of NO2are therefore necessary both to improve understanding
of the photochemistry of ozone, and to help monitor and
understand other critical atmospheric processes.
NO2is one of a family of odd nitrogen compounds in the
stratosphere with a complex diurnal variation in concentra-
tion. During daylight a balance is established between NO and
NO2concentrations through the reaction of the former with
ozone and the rapid photolysis and reaction with atomic
oxygen of the latter:
NO þ O3- NO2þ O2
NO2þ hn - NO þ O
NO2þ O - NO þ O2
Reaction (3) effectively determines the rate at which these
reactions destroy odd oxygen (O þ O3) since the first two
constitute a null cycle. At night, photolysis of NO2ceases and
all the odd oxygen in the stratosphere is converted to ozone—
so there is an almost-instantaneous change in the NO2
aSchool of Earth, Atmospheric and Environmental Sciences,
University of Manchester, UK M60 1QD. E-mail:
email@example.com; Fax: 44 161 306 3941;
Tel: 44 161 306 3931
bMet Office, FitzRoy Road UK Exeter EX1 3PB
cBritish Antarctic Survey, High Cross, Madingley Rd, Cambridge,
CB3 0ET, UK
dBelgian Institute for Space Aeronomy, Av Circulaire 3, B-1180
eCNRS Service d’Aeronomie, BP 3 Route des Gatines, 91370
This journal is ? c The Royal Society of Chemistry 2006J. Environ. Monit., 2006, 8, 353–361 | 353
PAPERwww.rsc.org/jem | Journal of Environmental Monitoring
concentration at twilight as the NO converts to NO2(in the
evening) or is created by photolysis (in the morning). At night,
NO2is converted first to NO3:
NO2þ O3- NO3þ O2
and then via a three-body reaction to the N2O5reservoir:
NO2þ NO3þ M - N2O5þ M (5)
This causes a build-up of N2O5during the night followed by a
slow release during the following day through photolysis. The
diurnal variation of NO2 therefore comprises a maximum
immediately after sunset, followed by a slow decrease through-
out the night and a sharp drop to a minimum at sunrise.
Thereafter, as N2O5 is photolysed NO2 builds up, finally
returning to its maximum as NO is converted to NO2 at
As well as the diurnal variation there is a seasonal variation
in stratospheric NO2at midlatitudes, due to the combined
effects of photochemistry and atmospheric transport. During
autumn and winter, NO2is converted to the long-term reser-
voir forms HNO3and ClNO3, either through gas-phase reac-
tions (e.g. NO2þ OH þ M - HNO3þ M) or hydrolysis of
N2O5 on aerosol particles. A minimum in NO2 occurs in
winter, sometimes exacerbated, particularly in the polar vor-
tex, by rapid conversion of N2O5to HNO3on polar strato-
spheric clouds. In spring and summer the reservoir species are
photolysed or destroyed by OH, re-generating NO2 which
therefore reaches a maximum around mid-summer. Although
these broad features in the NO2distribution are satisfactorily
explained by photochemical models, it seems that calculations
underestimate the amount of NO2,7especially below 25 km.8
This could be due to an incomplete understanding of NO2
chemistry on aerosol surfaces or uncertainty in gas-phase rate
Monitoring of stratospheric NO2from the ground or space
can be performed by absorption spectroscopy in the visible: an
absorption band around 450 nm introduces measurable opti-
cal depth to an extraterrestrial light source if the atmospheric
path is long. Such conditions apply to solar or lunar spectra
measured at the ground near twilight and to solar or stellar
occultation measurements from satellites. The first reported
measurements of stratospheric NO2by this technique were
those of Noxon10who used a scanning spectrophotometer to
make measurements for several years in the Colorado moun-
tains. His measurements clearly showed both the diurnal and
seasonal variations of stratospheric NO2concentration, and
also revealed the so-called ‘Noxon cliff’11,12—a sharp decrease
in NO2vertical columns at around 501N in winter due to
conversion of NOxto the reservoir species N2O5and HNO3in
the polar night.
The introduction of multi-diode arrays in the 1980s per-
mitted a considerable advance in the exploitation of the DOAS
technique, by enabling spectra to be recorded simultaneously
over a wide wavelength interval, without the problem of
intensity variations which affect scanning spectrometers. With
this technique, structure in the absorption spectrum of a
molecule, rather than the absolute amount, is used to measure
the path total.13,14The technique requires the absorption
species of interest to have a detailed structure within a fairly
narrow wavelength range, within about 10 nm. For strato-
spheric NO2, it involves dividing twilight spectra (when the
NO2optical depth is greatest) by a reference spectrum mea-
sured with high solar elevation, then filtering the resulting
ratio spectrum to leave high-frequency features resulting from
various important atmospheric absorbers, including ozone
and NO2. A particularly notable application of the technique
has been in measurements taken at high latitudes in late winter
and spring, when the chemistry of the lower stratosphere is
perturbed by polar stratospheric clouds.15–18
Stratospheric NO2 has been measured by a number of
satellite instruments since 1979, e.g. LIMS (Limb Infrared
Monitor of the Stratosphere19), SME (Solar Mesosphere Ex-
plorer20), SAGE-II (Stratospheric Aerosol and Gas Experi-
ment21), ISAMS (Improved Stratospheric and Mesospheric
Sounder22), HALOE (Halogen Occultation experiment23) and
POAM (Polar Ozone and Aerosol Measurement24). The most
extensive datasets have been obtained by SAGE-II and
HALOE, both of which use the solar occultation technique,
the first in the visible (448 and 453 nm) and the second in the
infra-red (5.26 and 6.25 mm). Drift in the optical components
of SAGE-II has caused considerable complications in deriving
the NO2product (although the latest version, 6.2, agrees better
with HALOE). Such drifts emphasise the need for high-quality
ground-based monitoring of NO2. Both instruments were also
affected below about 25 km by volcanic aerosol during the
period 1991–93, after the eruption of Mt Pinatubo. Strato-
spheric NO2columns have also been derived from the Global
Ozone Monitoring Experiment (GOME) instrument on ERS-
2, using the DOAS technique. This instrument has also
suffered its share of problems although it has been successfully
used to derive global fields of stratospheric NO2 vertical
columns and to delineate the position of the Noxon cliff.25
Nevertheless, space-based NO2 measurements are not yet
sufficiently advanced that they can be used reliably to derive
trends and interannual variations in NO2, so it is very im-
portant that high-quality ground-based measurements con-
tinue at a variety of locations world-wide.
A SAOZ UV-visible zenith-sky DOAS spectrometer has
been measuring both NO2 and ozone vertical columns at
Aberystwyth (52.41N, 4.21W) since March 1991. This site lies
on the western coast of the UK and normally experiences very
low pollution in the prevailing westerly airstream. Occasional
episodes of high tropospheric pollution are seen when air flows
from the east but the generally clean conditions make this a
good site for monitoring stratospheric NO2in northern mid-
latitudes. The SAOZ was occasionally operated at locations
other than Aberystwyth, as follows: Lerwick (601N, 11W) 2
Nov 1991–9 May 1992; Aberdeen (571N, 21W) 8 Feb–3 May
1994; Camborne (501N, 51W) 9–23 Sept 1994; OHPw (441N,
61E) 8–21 June 1996.
Here we present the results from March 1991 to September
2005 together with a discussion of the data quality and method
w Observatoire de Haute Provence.
354 | J. Environ. Monit., 2006, 8, 353–361This journal is ? c The Royal Society of Chemistry 2006
2.1.The SAOZ instrument
The SAOZ (Syste ` me d’Analyse par Observation Ze ´ nithale)
spectrometer was developed in the mid-1980s17to allow auto-
matic measurements of ozone and nitrogen dioxide at latitudes
up to the polar circle.26It consists of a commercial Jobin-
Yvon flat field spectrometer equipped with a concave holo-
graphic grating and a detector at ambient temperature com-
prising a Hamamatsu 512 or 1024 diode array (see below) with
a 25 or 50 mm entrance slit, respectively, giving a spectral
resolution of the order of 1 nm FWHM in the range 300–600
nm. A spectrometer of this design has been based at Aber-
ystwyth routinely measuring ozone and NO2 since March
1991.27The SAOZ observes the zenith sky with a field of view
of around 301, measuring light scattered downwards from a
range of altitudes. At twilight (Solar Zenith Angle, SZA B901)
the zenith sky is illuminated mainly by molecular scattering
from high altitudes. This light travels a long path through the
stratosphere, and by comparison a very short vertical path in
the troposphere. Absorption features in the spectra measured
at this time are therefore heavily biased towards the strato-
sphere. This makes zenith-sky spectroscopy ideal for distin-
guishing stratospheric NO2from boundary-layer pollution.
The spectrometer is housed in a dust- and water-proof con-
tainer with a quartz window to enable measurements from the
zenith sky. The instrument is controlled by a PC, recording
and analysing the spectra in real time. Measurements are
performed from sunrise to sunset while the Solar Zenith Angle
(SZA) is less then 941. The exposure time is adjusted auto-
matically between 0.1 s to 60 s in order to optimise the signal,
and the spectra are added to memory during a 60 s duty cycle.
The dark current is measured each time the duration of
exposure changes, and is then subtracted. Averages of ozone
and NO2morning and evening vertical columns are derived
from measurements taken between 871 and 911 SZA.
The original detector in the Aberystwyth SAOZ was a
Hamamatsu PCD (photoconductive device) 512 diode-array
detector. This was replaced in December 1992 with a Hama-
matsu NMOS (negative metal oxide semiconductor) 512
diode-array detector. The reason for this change was that
the NMOS detector has a thinner passivation layer, which
causes fewer interference effects than the thicker PCD detec-
tor. It also has a better quantum efficiency and lower dark
current, giving better signal to noise ratio. In May 1998 the
SAOZ was upgraded to a 1024 pixel detector with a change in
the grating from 200 mm?1to 360 mm?1. The slit width was
increased from 25 to 50 mm, corresponding to 0.9 and 1.2 nm
FWHM, resulting in an improvement of the sampling ratio
from 2.4 to 3.5 pixels over the slit width, reducing the inter-
polation error (see below).
The basic principle of the SAOZ data analysis is as follows:
each twilight spectrum is compared with a reference spectrum,
chosen at low zenith angle, to yield a slant column S—the
difference in NO2total column between the two light paths.
The vertical column total V corresponding to that spectrum
can then be derived using:
VðwÞ ¼SðwÞ þ R
where R is the residual amount of absorber in the reference
spectrum and AMF represents the air mass factor—the factor
by which the slant column at that particular zenith angle w
exceeds that of a vertical path. This equation is actually the
definition of the airmass factor, which is calculated using
radiative transfer models and appropriate profiles of ozone
and air density.28,29A standard set of AMFs is used in this
paper (Fig. 1). This set assumes a constant vertical distribution
of atmospheric absorbers and scatterers, and so does not vary
with season. Although Sarkissian et al.29quote an uncertainty
of 1.1% in the NO2AMF, more recent estimates suggest that
multiple scattering effects can increase the AMFs by 2–5%.30
For ozone, R may be evaluated by assuming that the ozone
concentration remains constant over the day when the refer-
ence spectrum is measured.31This method, however, is not
suitable for calculating the amount of NO2in the reference
spectrum since the concentration of NO2changes throughout
A Langley plot is a graph of slant column against AMF.
From such a plot for a given twilight period the slope gives the
vertical column density, and the intercept gives the amount of
absorber in the reference spectrum: V and R, respectively in
eqn (6), under the assumption that these do not change during
the twilight period. This assumption is not valid at dawn since,
as mentioned above, rapid changes of NO2concentrations
occur in the morning due to the photolysis of N2O5. In the
evening however, N2O5photolysis ceases and there is a period
between 801 and 901 SZA when the NO2vertical column is
effectively constant. A Langley plot during this period can
therefore be used to derive R[NO2], and is the method used in
The spectra measured by the SAOZ are recorded using real-
time programs provided by CNRS, and analysed using WinD-
OAS, the DOAS software suite developed at IASB.32The total
slant column of NO2is derived using the 432–470 nm spectral
window, taking into account absorption by ozone, O4and
H2O. In brief, the method uses non-linear least-squares fitting
of the log-ratio of two atmospheric spectra (measurement and
reference) to a set of molecular absorption cross-sections
measured in the laboratory at 220 K.33The actual algorithm
includes many advanced refinements, among which are
Fig. 1Air mass factors used in this work.
This journal is ? c The Royal Society of Chemistry 2006 J. Environ. Monit., 2006, 8, 353–361 | 355
capabilities for precise characterisation of both wavelength
calibration and spectral resolution of the instrument, and
calculation of the ring effect based on Raman scattering
modelling. In a special treatment designed for the SAOZ
instrument, the slit function is determined for each spectrum
and convoluted with the absorption cross-sections as part of
the DOAS fitting procedure.
The choice of reference spectrum was found to have a
substantial influence on the NO2analyses. In previous work
deriving ozone from the SAOZ,27a single reference spectrum
was used for periods when the instrument was fixed in one
location. For a long-term dataset this has the advantage that
any error in the amount of absorber in the reference spectrum
(R) does not affect any derived trend in the measurements.
However, for this method to work the detector must be rigidly
fixed with respect to the slit and diffraction grating—and in the
SAOZ it is not. This leads to a changing ‘pixel shift’—the
amount by which the measured spectrum must be shifted to fit
the reference—which degrades the accuracy of the NO2ana-
lyses and could introduce spurious ‘trends’. The obvious
solution is to use daily reference spectra, but in a location
where there is frequent rain and occasional episodes of pollu-
tion this leads to loss of good data. The compromise adopted
in this work was to use a monthly reference spectrum, and to
select that reference spectrum from a day near the middle of
the month which was free of pollution and recorded plenty of
signal (thus was free of rain). This approach led to the
retention of 1% more points in the dataset than with the daily
reference spectra. To check that the chosen reference spectrum
was a good one, spectra from 80–901 zenith angle on the
evening of that day were analysed with respect to the chosen
reference spectrum, and the Langley plot of slant column
against AMF examined. A correlation coefficient (r) greater
than 0.997 and a (negative) intercept within stipulated bounds
were required for an acceptable reference spectrum; the inter-
cept of this plot then gave the amount of NO2in the reference
Using monthly reference spectra means that the error in R
appears as a random error in any investigation of the long-
term trend in the data. The magnitude of this error was
estimated by analysing the 1995 dataset with 19 different
reference spectra, and comparing the results. This work was
carried out in 1999 with CNRS analysis software rather than
WinDOAS, giving a standard error of 7% attributable to the
choice of reference spectrum; it is unlikely to be substantially
different with WinDOAS.
For each twilight period (between 87 and 911) a weighted
average of the NO2vertical columns was calculated, using
error bars from the WinDOAS spectral fitting. (The number of
points per twilight ranged from 3 during the early years to
more than 20 in latter years, following improvements in soft-
ware and upgrades to the computer controlling the SAOZ.) To
eliminate bad data from the long-term dataset, the standard
deviation of the vertical columns during each twilight was used
as a quality indicator. Twilight periods where the standard
deviation was greater than 3 ? 1014cm?2were considered
unreliable and removed from the time series. A 3-point median
filter run through the afternoon and morning time series
separately removed most of the remaining bad data. Fig. 2
shows, for the data measured at Aberystwyth, the points
rejected as having excessive standard deviation and those
accepted after median filtering: the procedure is seen to be
remarkably successful in removing bad data.
In all, of the 5050 days between March 1991 and December
2004, SAOZ NO2measurements were obtained on 4300 of
them, i.e. 85% of the time. The fraction of days rejected by the
standard deviation criterion was 9% for the morning twilight
series and 12% for the evening twilights. These figures are
substantially better for the period of the 1024 pixel detector,
with measurements on 95% of all possible days and a rejection
to rejected points (standard deviation during twilight 43 ? 1014cm?2). A 3-point median filter has been applied to the accepted data. Solid black
line indicates changes in detector: the major change from 512 to 1024 pixels occurred in May 1998.
Removal of bad data points (Aberystwyth data up to end 2004 only). Blue and red symbols refer to data considered reliable, green points
356 | J. Environ. Monit., 2006, 8, 353–361This journal is ? c The Royal Society of Chemistry 2006
of 6% for morning and 9% for evening twilights. During this
period the SAOZ remained in Aberystwyth so data losses due
to transportation of the instrument did not occur.
2.3. Intercomparison with other instruments
As described above, the detector used within the SAOZ has
been changed twice during its time at Aberystwyth. The first
change merely improved the noise performance and had no
systematic effect on the measurements, but the introduction of
the 1024 pixel detector with the change in grating and slit
width was more far-reaching. It is important to understand
and quantify any effects these detector changes may have had
on the measurements, and to correct for any effects in order to
present a homogeneous dataset.
In June 1996 an NDSC (network for the detection of
stratospheric change) intercomparison took place at the Ob-
servatoire de Haute Provence (OHP) in Southern France
(43.9 1N, 5.71E). Sixteen instruments were involved, including
five SAOZs.34The Aberystwyth SAOZ was present, at that
time fitted with the NMOS 512 diode-array detector, as was
the OHP SAOZ with an NMOS 1024 diode-array detector
(referred to as instrument CNRS2 in Roscoe et al.34). The
latter is nominally identical to the 1024 pixel SAOZ currently
operating at Aberystwyth. The twilight data from this cam-
paign were all re-analysed using the same WINDOAS analysis
programs as the Aberystwyth data, using a single reference
spectrum from each instrument, to determine any systematic
offset between the 512 and 1024 pixel measurements.
Fig. 3 displays the results. There is a clear consistency
between the two datasets but also a tendency for the 512 pixel
measurements to be slightly higher. The weighted mean offsets
between the two data sets for morning and afternoon were
1.3 ? 0.9 ? 1014and 2.5 ? 1.2 ? 1014cm?2, respectively:
consistent with a value of 2 ? 1014cm?2? 50%. In percentage
terms this is 3.2% for the morning and 3.9% for the afternoon
data—about 3.5% on average, again with an uncertainty of
around half this amount. A further uncertainty of around the
same amount arises from the residual column (R in eqn (6)) in
each reference spectrum (see discussion of Fig. 7 below). This
intercomparison therefore cannot determine accurately any
systematic offset between the two detectors, but it does con-
firm that there is no glaring inconsistency between them. Note
the somewhat surprising increase in standard error for the
supposedly superior detector. We believe this to be due to the
asymmetrical slit function of the 1024 pixel SAOZ, compared
with the Gaussian shape of the 512 version.
The OHP intercomparison also provided an indication of
the relative accuracy of NO2monitoring instruments. In Fig.
12 of Roscoe et al.,34a comparison is presented of the mean
relative difference between each of the other instruments and
that of NIWA (New Zealand), chosen as the reference in this
case. These differences were corrected for the different cross-
sections used by each group. In this figure, the Aberystwyth
512 pixel SAOZ measured 16% less than NIWA and the OHP
1024 pixel SAOZ 8% less (a second 1024 pixel SAOZ operated
by the NILU group gave the same difference as the OHP
SAOZ). In other words, unlike the data in Fig. 3, the 512 pixel
SAOZ in Roscoe et al.’s paper showed 8% less NO2than its
1024 pixel counterpart. This is because the data presented in
Roscoe et al.34were analysed using earlier programs with a
cruder spectral fitting algorithm and a fixed slit width; the
refinements of WinDOAS produces better consistency be-
tween the different SAOZ instruments. Roscoe et al.34con-
cluded that there was a 1s scatter of 6% between the different
instruments present at the intercomparison and gave no
recommendation on the absolute accuracy of the measure-
From December 2000 to August 2003 a Brewer spectro-
photometer belonging to the UK Met Office was located in
Aberystwyth. This instrument measures NO2by direct solar
observation, rather than zenith sky, and is therefore more
sensitive to lower tropospheric pollution. During this period
the mean ratio SAOZ/Brewer for individual days was 1.01
(using afternoon Brewer data) with a standard deviation in the
ratio of 0.05. While this level of agreement is probably
flattering to the two instruments, it does confirm the high
quality of the Aberystwyth SAOZ dataset.
We note also here that the NO2values derived from the
SAOZ are significantly larger than the values derived from
GOME25. For 501N, Wenig et al.25report a maximum NO2
total column of 3.5 ? 1015cm?2in summer and a winter
minimum of 1 ? 1015cm?2. The solar zenith angle of the
GOME data varies with latitude and season so the two data
sets cannot be directly compared; nevertheless the daytime
NO2should be intermediate between the values at dawn and
dusk. However, the GOME summer maximum at 10 am is well
below that measured by the Aberystwyth SAOZ even at dawn
(Fig. 4). Further investigation is required to find the cause of
3. The Aberystwyth NO2time series 1991–2005
Presented in Fig. 4 is the time series of NO2measurements
from the Aberystwyth SAOZ spanning the period from 1991,
when the instrument was installed, to September 2005. The
time series also contains measurements from the same
detector) with OHP SAOZ (NMOS 1024 diode-array detector) vertical
column NO2measurements, when co-located at the 1996 OHP inter-
comparison. Error bars are 1s standard errors in the weighted mean
for each twilight, derived from the WinDOAS analysis.
Comparison of Aberystwyth SAOZ (NMOS 512 diode-array
This journal is ? c The Royal Society of Chemistry 2006 J. Environ. Monit., 2006, 8, 353–361 | 357
instrument while on campaigns at other locations, and has
been cleaned from the effects of pollution as described in
Standard errors in the mean of each twilight measurement
are shown in Fig. 5. These originate in the spectral fitting
procedure performed by WinDOAS, so larger values indicate
poorer spectral fits. For the 1024 pixel data (post-1998) the
afternoon errors are larger than in the morning, but for most
of the 512 pixel record there is little difference between the two.
When the errors are expressed as fractions (Fig. 6) this
correspondence is reversed: the 1024 pixel fractional errors
are the same for both twilight periods whereas the morning
512 data have relatively larger errors.
Clearly, there is a strong seasonal variation in the fractional
errors—caused by the seasonal variation in NO2itself. For
most of the year errors are o3%, but in the depth of winter
the errors increase to 410%. It is more illuminating therefore
to examine the absolute errors. These show a distinct variation
over the period of measurement with a reduction as expected
when the NMOS-512 detector was installed and a general
consistency during the 1024 pixel period. During 1994–1996
errors from the 512 pixel detector were smaller than during the
1024 period, and also very consistent. By contrast, when the
SAOZ returned from the OHP intercomparison in July 1996
there was a sharp increase in the errors, which subsided
somewhat in 1997 but returned in early 1998. During this
601N, 11W; Aberdeen, 571N, 21W; Camborne, 501N, 51W). Black line is as Fig. 2.
NO2vertical columns measured at Aberystwyth 1991–2005, including measurements made on campaigns at three other locations (Lerwick,
Fig. 5Standard errors in the mean value for each twilight. Black line is as Fig. 2.
358 | J. Environ. Monit., 2006, 8, 353–361This journal is ? c The Royal Society of Chemistry 2006
period problems were experienced with the instrument—ana-
lyses with a single reference spectrum revealed marked changes
in pixel shift over time suggesting a mechanical instability in
the detector, despite the SAOZ not being moved until the
upgrade in 1998. (These shifts are the reason a monthly
reference spectrum was used in the analyses presented in this
paper.) The sudden improvement in errors in April 1997 was
the result of refocussing the instrument. Nevertheless, the
quality of the data between July 1996 and April 1998 is not
as good as for the rest of the time series.
In order to focus more on interannual and long-term
changes in the NO2column, we have fitted to the Aberystwyth
data from 1994 onwards an annual and semiannual sine wave:
NO2ðdÞ ¼ a0? a1cos
d þ off1
d þ off2
using the coefficients in Table 1. The 512 pixel data have been
increased by 2 ? 1014cm?2, following the results of the OHP
intercomparison; although the uncertainty in this is large the
value itself is small and does not have a major impact on the
time series. The evening twilight results are shown in Fig. 7,
after filtering with an 11 point median filter to reduce the
scatter, and again with a 100 day Gaussian kernel to produce a
smooth curve. (Morning twilights showed similar variations
but with more scatter.) There is a clear tendency in Fig. 7 for
monthly data to cluster, emphasising that the uncertainty in
determining R (eqn (6)) is a significant source of error in this
technique. From examination of Fig. 7 this error (1s value)
can be estimated as B?3 ? 1014cm?2.
The most obvious feature in Fig. 7 is the sharp relative
reduction in NO2 in late 1991, caused by the eruption of
Mount Pinatubo (15.11N, 120.41E) in June of that year.
Aerosol from this eruption spread around the globe, reaching
Europe within 3 months, with the quantity of aerosol reaching
a maximum in early 1992.35Most of the reduction caused by
the volcano is due to a real change in the NO2column, with a
small component (no more than a few percent26) due to a
decrease in the air mass factor caused by aerosol scattering
from high altitudes. Data for Lerwick and Aberdeen are
included in Fig. 7 for illustration only: both stations are
affected by the polar vortex in winter and the substantial
decrease in observed NO2 is largely due to the change in
location. We cannot therefore infer from this dataset what
the NO2reduction due to Mt. Pinatubo was at Lerwick or
Aberystwyth in the winter of 1991–2. However, it is clear that
at Lerwick the diurnal variation in NO2columns was absent—
in contrast to the rest of the dataset. This feature of the
Pinatubo aerosol cloud was pointed out by Goutail et al.
Occasionally, in winter, the polar vortex extends southward
over the UK. As expected on such occasions a reduction in the
NO2column is observed by the SAOZ, of around 0.7 ? 1015
cm?2. Overall in winter there is a statistically significant
anticorrelation of NO2with potential vorticity at 475 K but
the correlation coefficient is only 0.16 so dynamical variability
explains only a small part of the variance in NO2.
From the summer of 1992, when the SAOZ returned to
Aberystwyth, to early 1995 the NO2 columns increase by
around 4 ? 10?14cm?2, or 10%; note that the 1995 values
Fig. 6As Fig. 5 but expressed as a fraction of the average column.
Coefficients of annual and semiannual oscillation fitted to
Amplitude of first harmonica
Offset of first harmonic, days
Amplitude of second harmonica
Offset of second harmonic, days
This journal is ? c The Royal Society of Chemistry 2006 J. Environ. Monit., 2006, 8, 353–361 | 359
are very similar to the pre-Pinatubo spring and summer of
1991. There is a step in the data of around 2 ? 1014cm?2when
the new detector was fitted in May 1998. It is unlikely that this
step is real and it would be possible to reduce the 512 pixel
series by this amount to remove it; the step is within the
uncertainty of the offset derived from the OHP intercompar-
ison. However, we prefer to leave the data set as it stands to
make the point that we cannot derive a long-term trend in NO2
across the whole dataset without making an arbitrary assump-
tion about the change in detector (see the last section of this
paper for a discussion of trends).
No such problems have plagued the data in recent years yet
there are interesting interannual variations in NO2. From 1998
to 2001 there is little departure from the model, consistent with
the data from New Zealand,2where the NO2annual cycle is
consistent for 1996–99. In 2002–05, by contrast, a distinct
wave-like variation appears in the columns with an amplitude
(peak–peak) of 7 ? 1014cm?2and an approximately biennial
period. In percentage terms this is large—the peak in the
summer of 2002 and 2004 corresponds to þ7% of the mea-
sured values, but the minimum in winter 2003 and mid-2005 is
nearer to ?10%. (The 512 pixel data show a similar variation
from 1995–97 but because of the problems with the instrument
this may not be real.)
We have presented in this paper fourteen years’ measurements
of stratospheric NO2total vertical columns from the Aberyst-
wyth SAOZ zenith-sky visible spectrometer. The instrument
spent most of its time at Aberystwyth with occasional excur-
sions to other locations, notably Lerwick in winter 1991–2 and
Aberdeen in the early months of 1994. Data have been analysed
with the WinDOAS analysis program developed at IASB, with
low-temperature high-resolution NO2cross-sections and fitting
a slit function to each spectrum. From March 1991 to April
1998 the SAOZ used a Hamamatsu 512 pixel diode-array
detector, which was upgraded to 1024 pixels in May 1998.
The choice of reference spectrum is crucial for the accuracy
of the WinDOAS analysis. Aberystwyth is a coastal location,
experiencing frequent rain and cloud, and it was found that a
reference spectrum taken in clear conditions (or at least not
raining) was necessary to derive NO2columns with minimal
standard error. The use of a single reference spectrum for
periods when the SAOZ was physically undisturbed had been
found to work well for ozone27but was not good enough for
NO2because small pixel shifts between reference and mea-
sured spectra degraded the analysis, particularly for the 512
pixel data. (Pixel shifts result from thermal expansion of the
optical components over a diurnal and annual timescale, and
from mechanical disturbance to the instrument.) Use of daily
reference spectra led to the occurrence of poor measurements
due to the effects of rain and thick cloud. The method used
here was to identify a reference spectrum from a clear day near
the middle of each month, and to use this to analyse that
month’s data. As well as producing a cleaner dataset, this has
the effect of converting the uncertainty in the residual column
amount R (eqn (6)) to a random error in a long time series. R
itself was derived from a Langley plot of slant column versus
air mass factor between zenith angles 801 and 901 on the day
the reference spectrum was measured. It was found that the
standard deviation of the vertical columns measured during
each twilight period provided a sufficiently robust criterion to
remove bad or polluted measurements: if the standard devia-
tion exceeded 3 ? 1014cm?2that twilight was discarded from
evening twilights only, 1991–2005. An 11-point median filter has been applied to the data, and values prior to the detector upgrade in 1998 reduced
by 0.2 ? 1015cm?2. Note the clustering of points caused by the use of monthly reference spectra; the scatter of these clusters indicates the
uncertainty in the residual column determination. Thick black line is a smoothed (100 day Gaussian filter) version of the blue points; thin magenta
line denotes changes of detector. Periods when the SAOZ was in Lerwick and Aberdeen marked with red squares.
Blue points: differences between measured NO2total columns and statistical model incorporating annual and semiannual variations,
360 | J. Environ. Monit., 2006, 8, 353–361 This journal is ? c The Royal Society of Chemistry 2006
There is a very distinct diurnal and annual variation in NO2
total columns, which are well-known features of odd nitrogen
chemistry in the stratosphere. To focus more on interannual
and long-term variations in NO2, a seasonal variation compris-
ing an annual and semi-annual component was fitted to the
morning and evening twilight separately (Table 1) from 1994 to
the present. This statistical model yielded average NO2columns
of 4.08 ? 1015cm?2and 2.68 ? 1015cm?2for the evening and
morning twilights, respectively, with a corresponding annual
amplitude of ?2.08 ? 1015cm?2and ?1.50 ? 1015cm?2.
Since 1998 the instrument has been stable, and the data can
be investigated for trends. A linear regression to the departures
from the model from 1998 to 2004 (blue points in Fig. 7) yields
a trend of 6.0 ? 0.5% per decade, which is consistent with the
results of Liley et al. (2000)2and Rinsland et al. (2003).4
However, it is clear that the time series is not statistically
stationary, for during the last three years an interannual
of B?8%, in contrast to the very steady measurements of
1998–2002. Determination of trends over the whole observa-
tion period is hampered by the change in detector: a linear
regression to the data in Fig. 7 for March–August 1991, Jan
1995–May 1996 and May 1998–December 2004 (i.e. omitting
data during the Pinatubo period and when the instrument was
not performing reliably) in fact yields a trend of ?1.0% ?
0.6% per decade. However, if the step which is evident at the
change of detector in Fig. 7 is removed from the data (i.e. the
512 pixel data are reduced by a further 2 ? 1014cm?2) the
deduced trend is 4.4% ? 0.3%. We therefore conclude that the
trends revealed by the Aberystwyth SAOZ agree with those
reported elsewhere for the period 1998–2004 and are not
inconsistent with them over the whole period of the dataset.
We thank the UK Natural Environment Research Council for
supporting this work, and David Moore and the UK Met
Office for loan of the Brewer spectrophotometer.
1 R. R. Garcia and S. Solomon, J. Geophys. Res., 1994, 99, 12937–
2 J. B. Liley, P. V. Johnston, R. L. McKenzie, A. J. Thomas and I. S.
Boyd, J. Geophys. Res., 2000, 105, 11633–11640.
3 E. Mahieu, R. Zander, P. Demoulin, M. de Mazie ` re, F. Me ´ len, C.
Servais, G. Roland, L. Delbouille, J. Poels and R. Blomme, in Air
Pollution Research Report no. 73, EUR 19340, ed N. R. P. Harris,
M. Guirlet and G. T. Amanatidis, 2000, pp. 99–102.
4 C. P. Rinsland, D. K. Weisenstein, M. K. W. Ko, C. J. Scott, L. S.
Chiou, E. Mahieu, R. Zander and P. Demoulin, J. Geophys. Res.,
2003, 108, 4437.
5 D. J. Fish, H. K. Roscoe and P. V. Johnston, Geophys. Res. Lett.,
2000, 27, 3313–3316.
6 C. A. McLinden, S. C. Olsen, M. J. Prather and J. B. Liley, J.
Geophys. Res., 2001, 106, 27787–27793.
7 M. Y. Danilin, J. M. Rodriguez, W. J. Hu, M. K. W. Ko, D. K.
Weisenstein, J. B. Kumer, J. L. Mergenthaler, J. M. Russell, M.
Koike, G. K. Yue, N. B. Jones and P. V. Johnston, J. Geophys.
Res., 1999, 104, 8247–8262.
8 S. Payan, C. Camy-Peyret, P. Jeseck, T. Hawat, M. Pirre, J. B.
Renard, C. Robert, F. Lefe ` vre, H. Kanzawa and Y. Sasano, J.
Geophys. Res., 1999, 104, 21585–21593.
9 G. B. Osterman, B. Sen, G. C. Toon, R. J. Salawitch, J. J.
Margitan, J. F. Blavier, D. W. Fahey and R. S. Gao, Geophys.
Res. Lett., 1999, 26, 1157–1160.
10 J. F. Noxon, Science, 1975, 189, 547.
11 J. F. Noxon, E. C. Whipple and R. S. Hyde, J. Geophys. Res., 1979,
12 J. F. Noxon, J. Geophys. Res., 1979, 84, 5067–5076.
13 M. Q. Syed and A. W. Harrison, Can. J. Phys., 1980, 58, 788–802.
14 U. Platt, D. Perner and H. W. Patz, J. Geophys. Res., 1979, 84,
15 G. H. Mount, R. W. Sanders, A. L. Schmeltekopf and S. Solomon,
J. Geophys. Res., 1987, 92, 8320–8328.
16 J. G. Keys and P. V. Johnston, Geophys. Res. Lett., 1988, 15,
17 J.-P. Pommereau and F. Goutail, Geophys. Res. Lett., 1988, 15,
18 S. Solomon and J. G. Keys, J. Geophys. Res., 1992, 97, 7971–7978.
19 J. M. Russell, J. C. Gille, E. E. Remsberg, L. L. Gordley, P. L.
Bailey, S. R. Drayson, H. Fischer, A. Girard, J. E. Harries and W.
F. J. Evans, J. Geophys. Res., 1984, 89, 5099–5107.
20 G. H. Mount, D. W. Rusch, J. F. Noxon, J. M. Zawodny and
C. A. Barth, J. Geophys. Res., 1984, 89, 1327–1340.
21 D. M. Cunnold, J. M. Zawodny, W. P. Chu, J. P. Pommereau, F.
Goutail, J. Lenoble, M. P. McCormick, R. E. Veiga, D. Murcray,
N. Iwagami, K. Shibasaki, P. C. Simon and W. Peetermans, J.
Geophys. Res., 1991, 96, 12913–12925.
22 W. J. Reburn, J. J. Remedios, P. E. Morris, C. D. Rodgers, F. W.
Taylor, B. J. Kerridge, R. J. Knight, J. Ballard, J. B. Kumer and S.
T. Massie, J. Geophys. Res., 1996, 101, 9873–9895.
23 L. L. Gordley, J. M. Russell, L. J. Mickley, J. E. Frederick, J. H.
Park, K. A. Stone, G. M. Beaver, J. M. McInerney, L. E. Deaver,
G. C. Toon, F. J. Murcray, R. D. Blatherwick, M. R. Gunson, J. P.
D. Abbatt, R. L. Mauldin, G. H. Mount, B. Sen and J. F. Blavier,
J. Geophys. Res., 1996, 101, 10241–10266.
24 C. E. Randall, D. W. Rusch, R. M. Bevilacqua, K. W. Hoppel and
J. D. Lumpe, J. Geophys. Res., 1998, 103, 28361–28371.
25 M. Wenig, S. Ku ¨ hl, S. Beierle, E. Bucsela, B. Ja ¨ hne, U. Platt, J.
Gleason and T. Wagner, J. Geophys. Res., 2004, 109, D04315.
26 F. Goutail, J.-P. Pommereau and A. Sarkissian, Geophys. Res.
Lett., 1994, 21, 1371–740.
27 A. C. Green, L. M. Bartlett and G. Vaughan, J. Quant. Spectrosc.
Radiat. Transfer, 2001, 69, 231–243.
28 S. Solomon, A. L. Schmeltekopf and R. W. Sanders, J. Geophys.
Res., 1987, 92, 8311–8319.
29 A. Sarkissian, H. K. Roscoe, D. J. Fish, M. Van Roozendael, M.
Gil, H. B. Chen, P. Wang, J.-P. Pommereau and J. Lenoble,
Geophys. Res. Lett., 1995, 22, 1113–1116.
30 M. R. Bassford, C. A. McLinden and K. Strong, J. Quant.
Spectrosc. Radiat. Transfer, 2001, 68, 657–677.
31 G. Vaughan, H. K. Roscoe, L. M. Bartlett, F. M. O’Connor, A.
Sarkissian, M. Van Roozendael, J.-C. Lambert, P. C. Simon, K.
Karlsen, B. A. K. Høiskar, D. J. Fish, R. L. Jones, R. A. Fresh-
water, J.-P. Pommereau, F. Goutail, S. B. Andersen, D. G. Drew,
P. A. Hughes, D. Moore, J. Mellqvist, E. Hegels, T. Klupfel,
F. Erle, K. Pfeilsticker and U. Platt, J. Geophys. Res., 1997, 102,
32 M. Van Roozendael, M. De Mazie ` re and P. C. Simon, J. Quant.
Spectrosc. Radiat. Transfer, 1994, 52, 231–240.
33 A. C. Vandaele, C. Hermans, P. C. Simon, M. Carleer, R. Colin, S.
Fally, M. F. Merienne, A. Jenouvrier and B. Coquart, J. Quant.
Spectrosc. Radiat. Transfer, 1998, 59, 171–184.
34 H. K. Roscoe, P. V. Johnston, M. Van Roozendael, A. Richter, A.
Sarkissian, J. Roscoe, K. E. Preston, J. C. Lambert, C. Hermans,
W. Decuyper, S. Dzienus, T. Winterrath, J. Burrows, F. Goutail, J.
P. Pommereau, E. D’Almeida, J. Hottier, C. Coureul, R. Didier, I.
Pundt, L. M. Bartlett, C. T. McElroy, J. E. Kerr, A. Elokhov, G.
Giovanelli, F. Ravegnani, M. Premuda, I. Kostadinov, F. Erle, T.
Wagner, K. Pfeilsticker, M. Kenntner, L. C. Marquard, M. Gil, O.
Puentedura, M. Yela, D. W. Arlander, B. A. K. Hoiskar, C. W.
Tellefsen, K. K. Tornkvist, B. Heese, R. L. Jones, S. R. Aliwell and
R. A. Freshwater, J. Atmos. Chem., 1999, 32, 281–314.
35 R. Neuber, G. Beyerle, G. Fiocco, A. di Sarra, K. H. Fricke, C.
David, S. Godin, B. M. Knudsen, L. Stefanutti, G. Vaughan and
J.-P. Wolf, Geophys. Res. Lett., 1994, 21, 1283–128.
This journal is ? c The Royal Society of Chemistry 2006J. Environ. Monit., 2006, 8, 353–361 | 361