Atmos. Chem. Phys., 8, 1483 1499, 2008
www.atmos-chem-phys.net/8/1483/2008/
' Author(s) 2008. This work is distributed under
the Creative Commons Attribution 3.0 License.
Atmospheric
Chemistry
and Physics
Validation of ACE-FTS satellite data in the upper troposphere/lower
stratosphere (UTLS) using non-coincident measurements
M. I. Hegglin1, C. D. Boone2, G. L. Manney3,4, T. G. Shepherd1, K. A. Walker1,2, P. F. Bernath2,5, W. H. Daffer6,
P. Hoor7, and C. Schiller8
1Department of Physics, University of Toronto, Toronto, Canada
2Department of Chemistry, University of Waterloo, Waterloo, Canada
3Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA
4New Mexico Institute of Mining and Technology, Socorro, New Mexico, USA
5Department of Chemistry, University of York, York, UK
6Columbus Technologies Inc., Pasadena, California, USA
7Max Planck Institute for Chemistry, Air Chemistry, Mainz, Germany
8
Institute for Chemistry and Dynamics of the Geosphere 1: Stratosphere, Research Centre J¤ulich GmbH, J¤ulich, Germany
Received: 27 August 2007 Published in Atmos. Chem. Phys. Discuss.: 26 September 2007
Revised: 3 January 2008 Accepted: 12 February 2008 Published: 13 March 2008
Abstract. CO, O3, and H2O data in the upper tropo-
sphere/lower stratosphere (UTLS) measured by the Atmo-
spheric Chemistry Experiment Fourier Transform Spectrom-
eter (ACE-FTS) on Canada’s SCISAT-1 satellite are vali-
dated using aircraft and ozonesonde measurements. In the
UTLS, validation of chemical trace gas measurements is a
challenging task due to small-scale variability in the tracer
elds, strong gradients of the tracers across the tropopause,
and scarcity of measurements suitable for validation pur-
poses. Validation based on coincidences therefore suffers
from geophysical noise. Two alternative methods for the
validation of satellite data are introduced, which avoid the
usual need for coincident measurements: tracer-tracer cor-
relations, and vertical tracer pro les relative to tropopause
height. Both are increasingly being used for model validation
as they strongly suppress geophysical variability and thereby
provide an instantaneous climatology . This allows com-
parison of measurements between non-coincident data sets
which yields information about the precision and a statisti-
cally meaningful error-assessment of the ACE-FTS satellite
data in the UTLS. By de ning a trade-off factor, we show
that the measurement errors can be reduced by including
more measurements obtained over a wider longitude range
into the comparison, despite the increased geophysical vari-
ability. Applying the methods then yields the following up-
per bounds to the relative differences in the mean found be-
Correspondence to: M. I. Hegglin
(michaela@atmosp.physics.utoronto.ca)
tween the ACE-FTS and SPURT aircraft measurements in
the upper troposphere (UT) and lower stratosphere (LS), re-
spectively: for CO ±9% and ±12%, for H2O ±30% and
±18%, and for O3 ±25% and ±19%. The relative differ-
ences for O3 can be narrowed down by using a larger dataset
obtained from ozonesondes, yielding a high bias in the ACE-
FTS measurements of 18% in the UT and relative differences
of ±8% for measurements in the LS. When taking into ac-
count the smearing effect of the vertically limited spacing
between measurements of the ACE-FTS instrument, the rel-
ative differences decrease by 5 15% around the tropopause,
suggesting a vertical resolution of the ACE-FTS in the UTLS
of around 1 km. The ACE-FTS hence offers unprecedented
precision and vertical resolution for a satellite instrument,
which will allow a new global perspective on UTLS tracer
distributions.
1 Introduction
The upper troposphere/lower stratosphere (UTLS) has re-
cently attracted major research interest in atmospheric sci-
ence due to its key role in chemistry-climate coupling. In
order to characterize UTLS tracer distributions and to detect
future changes, tracer measurements with global coverage
are needed. Satellites are the only means by which this task
can be accomplished, however, their capability to measure
accurately in the UTLS is limited, and the validation of the
measurements dif cult.
Published by Copernicus Publications on behalf of the European Geosciences Union.

1484 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
PVU
Fig. 1. Vertical cross section of ECMWF potential vortic-
ity (colour coded) from equator to North Pole at 6◦ W on
10 November 2001 at 06:00 UTC. White thick line shows the
dynamical tropopause (2 PVU, with 1 PVU=1 potential vorticity
unit=10−6 m2 s−1K kg−1), white dashed line the 380 K isentrope.
Green lines indicate the location of two independent pro les that
ful ll the spatial coincidence criteria of being taken no further apart
than 500 km.
These limitations arise from the dynamical and chem-
ical structure of the UTLS. Dynamical variability in the
tropopause region induced by Rossby wave activity is high,
and length scales of the associated features in the tracer elds
are small less than 1 km in the vertical, and 100 km in the
horizontal. Tracer mixing ratios also exhibit a strong gradi-
ent across the tropopause because it acts as a transport bar-
rier (Pan et al., 2004; Hoor et al., 2004; Hegglin et al., 2006
and references therein). Ultimately, the remote sensing tech-
nique used determines the achievable vertical and horizon-
tal resolution of the measurements and hence the capabil-
ity of the instrument to resolve the given geophysical small-
scale variability. The Atmospheric Chemistry Experiment
Fourier Transform Spectrometer (ACE-FTS) on Canada’s
SCISAT-1 satellite (Bernath et al., 2005) has provided accu-
rate measurements of numerous chemical species throughout
the stratosphere and into the UT since February 2004, with
a vertical spacing between measurements of less than 1 km
at the lowest retrieval altitudes. The measurements suggest a
high potential for studies related to the UTLS.
In the stratosphere and mesosphere, the ACE-FTS satellite
measurements are being validated by comparison to balloon-
borne or independent satellite data which are approximately
coincident in time and space (see other publications in this
special issue on ACE validation). Coincidence criteria are
de ned in various ways, but typically the measurements have
to be taken within several hours and at locations no further
apart than around 500 km. However, in the case of occulta-
Fig. 2. Left panel: Vertical pro les of CO mixing ratios taken
during ascent (black) and descent (red) of the SPURT aircraft on
10 November 2001 from and to the airport Hohn (Northern Ger-
many). The pro les were taken approximately 10 h apart. Right
panel: Relative difference of the measured pro les. Gray shading
indicates an error range of ±10%.
tion sounders, it can be dif cult to nd a suf cient number of
coincident measurements for a statistically meaningful val-
idation (Walker et al., 2005). Furthermore, this validation
method does not account for differences in the tracer mix-
ing ratios which are produced by the geophysical variability
found within the de ned time and length scales.
The use of coincident measurements is especially an is-
sue in the UTLS, where geophysical variability is large and
a strong gradient in chemical tracers is found across the
tropopause. Figure 1 provides a Northern Hemisphere cross
section of ECMWF potential vorticity at 6◦ W on 10 Novem-
ber 2001 at 06:00 UTC. The dynamical tropopause (white
line) shows strong undulations and even pro les taken no fur-
ther apart than 500 km exhibit large differences in tropopause
height (green lines). Figure 2 shows two vertical pro les
of CO mixing ratios measured in-situ by an aircraft instru-
ment during the same dynamical situation and taken within
8 h and a distance of 400 km. As can be seen, the apparent
error derived for the measurements in the tropopause region
is as large as 50%. However, this is due to the geophysi-
cal variability and not a measurement error (the pro les are
taken by the same instrument). Further complicating a sta-
tistically meaningful comparison is the sharp tracer gradient
at the tropopause, which causes sampling errors to be non-
normally distributed.
On the other hand, sparse data sets are increasingly being
used for the validation of chemistry transport and chemistry
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1485
climate models, through diagnostics that provide instan-
taneous climatologies and reveal characteristic features in
tracer distributions (Douglass et al., 2003; Sankey and
Shepherd, 2003; Hegglin and Shepherd, 2007; Strahan et
al., 2007). In the UTLS, such characteristic features are
found when moving from geometric altitude coordinates into
tracer-tracer space or tropopause coordinates. These trans-
formations strongly reduce geophysical variability and help
in assessing the quality and vertical information content of
the model data. The same approach can therefore be consid-
ered for the validation of new satellite data.
In this study, we investigate the use of tracer-tracer corre-
lations and vertical tracer pro les relative to the tropopause
height for the validation of the ACE-FTS CO, O3, and H2O
measurements in the UTLS. Section 2 provides the descrip-
tion of the data sets used. In Sect. 3, the new validation meth-
ods are introduced, and their strengths and weaknesses dis-
cussed. We assess whether spatial and temporal variations
in the climatologies are acceptable for identifying vari-
ous types of errors. The methods are then used in Sect. 4
to validate the ACE-FTS measurements using aircraft and
ozonesonde data. A summary is given in Sect. 5.
2 Data description
2.1 ACE-FTS satellite data
The Atmospheric Chemistry Experiment Fourier Trans-
form Spectrometer (ACE-FTS) is the primary instrument on
SCISAT-1, a Canadian-led satellite mission for remote sens-
ing of the Earth’s atmosphere. ACE-FTS features high res-
olution (0.02 cm−1) and broad spectral coverage in the in-
frared (750 to 4400 cm−1). The instrument operates almost
exclusively in solar occultation mode (Bernath et al., 2005).
The SCISAT-1 satellite was launched into low Earth circu-
lar orbit (650 km) with high inclination (74◦) on 12 August
2003. In solar occultation mode, this orbit provides season-
ally varying coverage of the globe, with an emphasis on mid-
latitudes and polar regions. Up to 30 occultation events (sun-
rises or sunsets viewed by the orbiting satellite) occur per
calendar day. Science operations for the ACE-FTS began in
February 2004. Retrievals for the ACE-FTS employ a non-
linear least squares global- t approach (Boone et al., 2005).
The instrument has a 1.25 mrad input aperture, which sub-
tends an altitude range of 3 4 km at the tangent point (the
point of closest approach to the Earth for a solar ray mea-
sured by the instrument). However the ACE-FTS instrument
collects measurements every 2 s, which can lead to signi -
cant oversampling. The altitude spacing between measure-
ments varies over the course of the year, governed primarily
by the beta angle (the angle between the satellite orbit plane
and the Earth-Sun vector) corresponding to the occultation.
The rate of change of the tangent height of each measurement
within an occultation decreases with increasing beta angle,
Fig. 3. Probability density function of the vertical spacing between
the ACE-FTS measurements (dz) as a function of retrieval altitude.
The vertical spacing in the tropopause region (between approxi-
mately 8 and 14 km) exhibits values that can be much lower than
1 km.
leading to higher vertical sampling for larger beta angles. At
low altitudes, refraction effects also impact the measurement
spacing. ACE-FTS results are provided on two altitude grids,
a 1-km grid common to all occultations and a retrieval grid
that varies from occultation to occultation. The retrieval grid
contains values at the measurement altitudes, unless there are
multiple measurements within a layer on the 1-km grid, in
which case it provides a single value at the center of the 1-km
grid layer. For this study, we use the measurements on the re-
trieval grid. Figure 3 shows the probability density function
of the vertical spacing between altitudes on the retrieval grid
as a function of height for all measurements between 5 and
50 km. The altitude spacing in the UTLS varies from about
3 km to less than 1 km. Part of the purpose of this paper is
to assess how much real vertical resolution there is in the re-
trieval grid.
In this study we focus on the validation of the ACE-FTS
version 2.2 H2O, CO, as well as the version 2.2 O 3 up-
date results in the UTLS between February 2004 and Jan-
uary 2007. The UTLS comprises the tropopause region, and
lies between approximately 5 and 15 km altitude or 500 and
100 hPa. The validation of the stratospheric and mesospheric
ACE-FTS version 2.2 data (with ozone updates) is published
along with the paper presented here in the special issue on
validation of ACE (for CO: Clerbaux et al. (2007), for O3:
Dupuy et al. (2008), for H2O: Carleer et al. (2008)). Ear-
lier comparisons of the version 1.0 ACE-FTS O3 data with
GOMOS, POAM III, and SAGE III satellite data (Fussen et
al., 2005; Walker et al., 2005) yielded good agreement be-
tween the data sets with differences mostly lower than 10%
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1486 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
between 15 and 45 km. With HALOE, an agreement of ±5%
between 15 and 35 km was found (McHugh et al., 2005). In
the same study, the ACE-FTS H2O data indicated a high bias
of around 20% at altitudes below 20 km. A rst validation of
the version 1.0 CO by Jin et al. (2005) using Odin satellite
data yielded excellent agreement, and a comprehensive anal-
yses of these data was provided by Clerbaux et al. (2005).
ACE-FTS version 2.1 level data were compared to the Aura-
MLS version 1.5 (Froidevaux et al., 2006), and Aura-MLS
version 2.2 satellite data (Lambert et al., 20071; Froidevaux
et al., 2008; Pumphrey et al., 20072). The MLS/ACE-FTS
differences for O3 in the lower stratosphere are within ±5%,
but increase with altitude. For H2O, the instruments show
good agreement, within ±5%, between 68 0.004 hPa. At the
lowest levels considered in this study (around 100 hPa) the
differences increase up to 30%. Differences in the CO mea-
surements are around 50% in the stratosphere, and 25% in
the mesosphere.
2.2 SPURT aircraft data
As a reference data set, we use in-situ high-resolution and
high-precision CO, O3, and H2O measurements from the
SPURT (German acronym for trace gas transport in the
tropopause region ) aircraft campaign. The ights were car-
ried out seasonally between November 2001 and July 2003
in the Northern Hemisphere over Europe and cover a lati-
tude range between 30◦ N and 80◦ N. For each season, ap-
proximately 32 ight hours or 24 000 data points were ob-
tained. An overview of the campaign can be found in Engel
et al. (2006). Detailed descriptions of the CO, O3, and H2O
measurements can be found in Hoor et al. (2004), Hegglin et
al. (2006), and Krebsbach et al. (2006), respectively.
2.3 WOUDC ozonesonde data
In order to corroborate the results obtained by the valida-
tion using the spatially limited aircraft measurements we ex-
tend our evaluation to the WOUDC ozonesonde data set.
The data set is available at the WOUDC Data Web page
(http://www.woudc.org/data e.html) and includes over 300
stations worldwide. We use temperature and ozone pro les
with a vertical resolution of 100-150 m obtained during the
years 2004 2007 and in a latitude range between 40 ◦ N and
60◦ N. The accuracy of ozonesonde observations is estimated
to be ±5% (e.g. SPARC, 1998).
1Lambert, A., Read, W. G., Livesey, N. J., et al.: Validation of
the Aura Microwave Limb Sounder middle atmosphere water va-
por and nitrous oxide measurements, J. Geophys. Res., submitted,
2007.
2Pumphrey, H. C., Filipiak, M. J., Livesey, N. J., et al.: Valida-
tion of middle-atmosphere carbon monoxide retrievals from MLS
on Aura, J. Geophys. Res., submitted, 2007.
2.4 Derived meteorological products
For each measurement of the different data sets, we derived
the thermal tropopause height according to the WMO def-
inition, i.e. the lowest level at which the lapse rate drops
to 2 K km−1 or less, and the average lapse rate between
this level and all higher levels within 2 km does not ex-
ceed 2 K km−1 (WMO, 1957). The tropopauses for the
ozonesonde data were calculated using simultaneously mea-
sured temperature pro les, and those for the ACE-FTS and
the SPURT data using the Goddard Earth Observing Sys-
tem Model, Version 4 (GEOS-4), by interpolation of the
model elds onto the exact measurement location in time
and space and by applying the Reichler et al. (2003) al-
gorithm. The GEOS-4 analyses are described by Bloom
et al. (2005). The GEOS-4 data used here are provided
on 55 hybrid (σ /pressure) model levels from the surface to
0.01 hPa. The horizontal grid is 1.0◦ latitude by 1.25◦ lon-
gitude. Six-hourly average elds are provided centered at
00:00, 06:00, 12:00 and 18:00 UTC. Besides the standard
meteorological variables, GEOS-4 products include an ex-
tensive set of elds from the model and assimilation sys-
tem, including PV calculated internally in the model. Fur-
ther information on the derived meteorological products can
be found in Manney et al. (2007).
2.5 Canadian Middle Atmosphere Model data
The Canadian Middle Atmosphere Model (CMAM) is used
to investigate the geophysical variability in the tracer dis-
tributions on different time and spatial scales, which poten-
tially in uences the reliability of the validation methods in-
troduced here. CMAM is an extension of the Canadian Cen-
tre for Climate Modelling and Analysis (CCCma) spectral
General Circulation Model into the lower thermosphere up
to 0.0006 hPa (around 100 km). CMAM includes a compre-
hensive representation of the relevant physical and chemical
processes in a fully interactive mode (Beagley et al., 1997;
de Grandpr·e et al., 2000). Data presented here correspond to
the 3-day output of instantaneous elds in the years 2000
2010 of a 150-year transient run from 1950 with CMAM
version 8, using model-generated sea-surface temperatures
and background aerosol forcing. Model elds are calculated
on a linear Gaussian transform grid with 32×64 grid points
in the horizontal, corresponding to a resolution of around
6◦×6◦, and 71 vertical levels. The vertical resolution in
the tropopause region is around 900 m, increasing to around
2 km in the upper stratosphere. A detailed comparison of
CMAM version 8 (run over 1960 2004 with observed sea
surface temperatures and volcanic aerosol forcing) with ob-
servations is provided as part of the model intercomparison
of Eyring et al. (2006).
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1487
3 New validation methods
The underlying idea of the validation methods introduced
here is based on the widely accepted use of characteris-
tic features found in tracer climatologies to test the simu-
lation of tracer distributions in chemistry climate or chem-
istry transport models (Douglass et al., 2003; Sankey and
Shepherd, 2003; Hegglin and Shepherd, 2007; Strahan et al.,
2007). These climatologies are often obtained from spatially
and temporally inhomogeneous tracer observations, yet are
considered to represent the full atmosphere. In the UTLS,
characteristic features in tracer distributions are found when
moving from geometric altitude coordinates into tracer-tracer
space or tropopause coordinates, as explained in more detail
below. Our hypothesis is that new observations, such as the
ACE-FTS data set, can be considered accurate and precise if
shown to be capable of resolving the characteristic features
found in such climatologies. Since this approach allows one
to include all measurements, not just the coincident ones, one
can get much better statistics. When using tracer-tracer cor-
relations in the lower stratosphere, for example, model errors
of the order of 10% can be detected (Hegglin and Shepherd,
2007).
3.1 Tracer-tracer correlations
Suf ciently long-lived species exhibit compact correlations
(Plumb and Ko, 1992), which reduce day-to-day variations
and provide an instantaneous climatology . Depending on
the lifetimes of the tracers used, compact correlations are
not necessarily linear; as in the case of the O3-CO and
O3-H2O correlations, they can exhibit a strong curvature
which can be used to identify the chemical transition be-
tween the troposphere and the stratosphere (Pan et al., 2007
and references therein). Apart from recent troposphere-to-
stratosphere transport events, which produce distinct and
nearly linear mixing lines (Fischer et al., 2000), the compact-
ness of these correlations is relatively high and their shape
distinct due to their strong dependence on the location of the
tropopause, so that they can be used to gauge the precision
of the ACE-FTS measurements.
3.2 Vertical pro les relative to tropopause height
Another method for reducing the effects of geophysical vari-
ability in UTLS tracer measurements, and for obtaining elds
suitable for comparison of non-coincident measurements, is
the use of tracer vertical pro les relative to the tropopause
height (i.e. in tropopause coordinates) (cf. Hoor et al., 2004;
Pan et al., 2004; Hegglin et al., 2006). These pro les are
more compact and show less scatter than data plotted in geo-
metric altitude (or potential temperature) coordinates. This
is shown in Fig. 4a and b for ozonesonde data obtained
during winter (DJF) and accumulated over the years 2004
2007. They furthermore reveal a sharp gradient between tro-
Fig. 4. Ozonesonde data obtained in Northern Hemisphere winters
2004-2007 between 40◦ N and 60◦ N. (a) In geometric altitude and
indicating the average height of the tropopause (gray solid line) ±
its standard deviation σ (dashed gray lines), (b) in tropopause coor-
dinates, and (c) plotted as means ±σ with the data from (a) in gray
plotted relative to the mean tropopause height (subscript dz geo),
and data from (b) in black (subscript dz tp). (d) Ratio between the
standard deviations of the two sample methods (σdz tp/σdz geo),
now depicted for all seasons.
pospheric and stratospheric tracer mixing ratios which can
be used to test the vertical resolution and information con-
tent of the ACE-FTS data. A recent study by Monahan
et al. (2007) used tropopause coordinates for the validation
of O3 data from the AIRS satellite using coincident mea-
surements. They showed that tropopause coordinates allow
the separation of the measurement errors between the tro-
posphere and the stratosphere, which led to a decrease and
increase in the error in lower stratospheric and upper tropo-
spheric measurements, respectively. We show here that this
validation method, when used in a climatological rather than
an instantaneous way, can be applied even to non-coincident
measurements, allowing a statistically meaningful error as-
sessment for sparse data sets.
In order to apply the tropopause coordinates to the ACE-
FTS and SPURT or ozonesonde data, we rst calculate the
distance from the thermal tropopause of each measurement
point. We then calculate the CO, O3, and H2O mean mixing
ratios (x) and their standard deviations (σ ) for each 1 km alti-
tude bin from −6 to +6 km relative to the tropopause height.
We nally derive the relative differences ( 1rel) between the
mean pro les of the ACE-FTS (subscript ACE) and the ref-
erence data set (subscript ref) using
1rel =
xACE − xref
0.5 × (xACE + xref)
, (1)
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1488 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
Fig. 5. Interannual variability in the seasonal means of vertical
ozone pro les in tropopause coordinates for the latitude range be-
tween 40◦ N and 60◦ N and all longitudes, calculated from (a) ten
years of CMAM data and (b) ozonesondes taken between 2004 and
2007.
and their relative uncertainties (δrel) calculated according to
the general rules of error propagation, and using the standard
errors (δ) of each mean value estimated by
δ = σ√N/b . (2)
Here, N is the number of measurements, and b is a factor to
account for the autocorrelation length of the measurements,
taken to be 30, 45, and 3 for ozonesonde, SPURT aircraft,
and ACE-FTS data, respectively. For the SPURT data set,
b corresponds to the autocorrelation length between mea-
surements along a ight track which is mainly determined
by the length scales generated by the prevailing dynamical
elds ( ≈350 km). For the ozonesondes and the ACE-FTS
pro les, b is determined by the maximum number of mea-
surements per pro le expected to end up in the same altitude
bin. Single pro les are assumed to be independent of each
other. Conservative values were chosen for these parameters.
3.3 Strengths and limitations of the validation methods
The use of instantaneous climatologies in the form of
tracer-tracer correlations or vertical pro les in tropopause
coordinates allows one to compare a much larger number
of measurements than using coincident measurements only.
This helps reduce the standard errors of the derived mean val-
ues, and hence allows a statistically more meaningful error
assessment. The downside of including more measurements
taken over larger geographical regions and longer time pe-
riods, however, is that the geophysical variability increases.
Furthermore the instantaneous climatology will itself vary,
which introduces a potential source of error if the two data
sets have different sampling. In the worst case, if these
spatial and temporal variations are larger than the measure-
ment errors, the real uncertainty in the measurement will
never be found. Hence the questions to be addressed in our
study are: how much daily geophysical variability can be
eliminated? Can the elimination of this variability compen-
sate for the inherent uncertainty when using climatologies
in tropopause coordinates? How many independent mea-
surements are needed to describe this climatology? We will
answer these questions by focusing on vertical pro les in
tropopause coordinates. However, the derived results will be
valid as well for the tracer-tracer correlations.
3.3.1 Daily variability and climatological uncertainty
We can estimate the reduction in geophysical variability by
comparing the standard deviations of the mean O3 values
calculated from the ozonesondes in each coordinate system
(Fig. 4c). The data in geometric altitude is plotted relative to
the mean tropopause height in order to allow direct compar-
ison of the two coordinate systems. Moving from geomet-
ric altitude to tropopause coordinates reduces the geophysi-
cal variability by up to 50% around the tropopause (Fig. 4d).
This accounts for the very large standard deviations found in
the error assessment for the ACE-FTS O3 at altitudes below
15 km (note the tropopause is found around 10 12 km) using
coincident measurements (Fig. 46 in Dupuy et al., 2008), and
implies that accounting for the daily geophysical variability
will greatly improve the validation. It is furthermore inter-
esting to see that the effectiveness of the tropopause coor-
dinate system in reducing geophysical variability decreases
with increasing distance from the tropopause, and reverses
sign beyond a certain distance. This evaluation reveals that
the in uence of the tropopause in shaping UTLS tracer dis-
tributions is limited to a certain depth which is determined by
the season. During winter and spring, variability is reduced
up to 5 km (4 km) above (below) the tropopause, but only
up to 2 km (3 km) in summer and autumn. Hence the use
of tropopause coordinates is bene cial only in this altitude
range.
The question about the inherent uncertainty of instanta-
neous climatologies using tropopause coordinates is more
dif cult to answer since it is in uenced by geophysical vari-
ability on various time and length scales. This cannot be
determined by observations alone since they are limited in
space and time, and might be affected by sampling biases.
However, we can use a model which provides information
about the full atmosphere (in our case CMAM), to estimate
the geophysical variability on the time and length scales
chosen. In order to test the capability of CMAM to cap-
ture geophysical variability appropriately, we compare the
interannual variability in the seasonal means obtained for
a latitude range between 40◦ N and 60◦ N derived from ten
years of CMAM data and from 3 years of ozonesonde data
(Fig. 5). The evaluation shows that the variability simulated
by CMAM lies around 10% for all altitude levels. The in-
terannual variability of the seasonal means derived from the
ozonesonde observations shows a similar range in the LS,
and a smaller one for the UT. This evaluation provides us
with some con dence that we can use CMAM to test sam-
pling requirements.
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1489
We now determine how many independent measurements
are needed to represent the mean pro les of the full atmo-
sphere, so that the error introduced by the spatial and tempo-
ral variability inherent in the climatology is acceptable. To
do this we use CMAM data obtained during a single winter
season (DJF) from latitudes between 40◦ N and 60◦ N (in-
cluding all longitudes). Datasets for 12 altitude levels be-
tween −6 and 6 km relative to the tropopause height (with
a bin size of 1 km) are generated, which will be referred to
henceforth as full sampling. From these datasets, we gen-
erate 1000 smaller ( subsampled ) datasets each of which
is composed of a different number of measurements ranging
from 1 to 1000. Means and standard deviations are computed
for each of the full sampling and the subsampled datasets.
The relative differences of the means between full sampling
and the subsampled datasets for each altitude level (colour
coded) are shown in the top panels of Fig. 6 (O3 in panel a,
H2O in panel b). The relative differences decrease with the
number of data points included in the subsample. They are
generally larger for lower (bluish colours) than for higher al-
titude levels (reddish colours), and larger for H2O than for
O3. The relative differences lie on average within 10% for
H2O, but within 5% for O3.
In order to show that the relative differences are decreasing
with increasing number of included measurements following
theoretical expectations, the 1000 subsampled datasets of
each level are regenerated 20 times using randomized data,
and their means and standard deviations are computed. The
relative differences between the means of these 20 subsam-
pled and the fully sampled datasets are shown in the lower
panels of Fig. 6 (gray shadings) for selected altitude levels.
The colored curves indicate the theoretically expected de-
crease in the relative differences, which is described by the
standard error of the mean (σ /
√
N ), where σ is the standard
deviation of the full sampling. The decrease is dependent on
the respective altitude level. The relative differences are seen
to be largest just below the tropopause, and are smaller in the
LS and in the troposphere. Also denoted are the number of
measurements (black asterisks) needed to reduce the relative
differences to ±5% for O3, and to ±15% for H2O. 50 100
independent measurements are needed for O3, and 80 120
for H2O, depending on the altitude level.
3.3.2 Trade-off factor
Given an available set of measurements, we now need to as-
sess whether it is bene cial to move to a wider longitude
range and/or add more years of data. Expressed in another
way, we need to know whether the trade-off between increas-
ing geophysical variability and adding more measurements
is helping to improve our error assessment. To this end we
compare the standard errors of the mean (σ /
√
N ), shown in
the previous section to describe the decrease in relative dif-
ferences, for CMAM datasets including different longitude
ranges and years. Examples of this evaluation are shown in
Fig. 6. Relative differences between the means of the full model
eld and the means derived from a speci ed number ( N) of in-
dependent data points plotted as a function of N for (a) O3, and
(b) H2O. Data used are restricted to the latitude range between
40◦ N and 60◦ N and DJF of one year. Upper panels: one cal-
culation for each altitude bin between −6 and +6 km above the
tropopause (1 km spacing, colour coded) with dashed/solid lines in-
dicating ±5% and ±10% relative difference. Lower panels: 20 cal-
culations using randomized data (in gray) for selected altitudes rel-
ative to the tropopause height (−5.5, −2.5, −0.5, 0.5, 2.5, 5.5 km,
colour coded) and offset by 75, 45, 15, −15, −45, and −75% for
O3, and by 250, 150, 50, −50, −150, −250% for H2O, respec-
tively. Dashed lines indicate in (a) ±5%, and (b) ±15% relative
difference. The coloured lines illustrate how the standard error of
the mean (±σ/
√
N , with σ being the standard deviation of the full
sampling) is decreasing with increasing number of measurements.
The black asterisks indicate the number of measurements needed to
decrease the standard error of the mean to ±5% and ±15% for O3
and H2O, respectively.
Fig. 7. The black curves denote the standard errors of the
mean computed using the σ for data obtained over one year
of DJF data between 40◦ N and 60◦ N, and restricted to lon-
gitudes between 0◦ E and 45◦ E. The gray curves are for the
full longitude range. The upper curves are for O3 at 3.5 km,
the lower curves for H2O at 1.5 km above the tropopause.
Red triangles indicate the number of available independent
SPURT measurements (N/b), and black and gray asterisks the
number of independent ACE-FTS measurements for the re-
stricted and full longitude ranges, respectively. Note that the
ozonesondes provide over 200 independent measurements
for both full and restricted longitude ranges and all altitude
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1490 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
Fig. 7. Standard errors of the mean (σ /
√
N ) as a function of number
of measurements (N) for O3 at 3.5 km (upper curves) and H2O at
1.5 km above the tropopause (lower curves, sign has been ipped)
and for two different CMAM datasets including DJF data from 5
years and a latitude range between 40◦ N and 60◦ N. Black lines
denotes the calculation for data restricted to a longitude range of
0◦ E 45 ◦ E, gray lines include measurements from all longitudes.
Black and gray asterisks denote the number of independent ACE-
FTS measurements (N/b) for the restricted and full longitude ranges
during DJF and between 2004 2007. Red triangles denote the num-
ber of independent SPURT measurements. Red arrow indicates the
gain in number of measurements, blue arrow the decrease in rela-
tive difference when moving from the restricted to the full longitude
range.
levels, hence can be regarded as a perfect validation data
set (the data points lie outside of the plotted range). When in-
cluding more longitudes (moving from the black to the gray
curves), the standard errors in the mean (i.e. the geophysi-
cal variability) for a given number of measurements increase
for both tracers, but less so for O3 than for H2O. However,
when including more longitudes, more ACE-FTS measure-
ments are available (the gain is denoted by the red arrows),
and the relative differences are reduced (the gain is denoted
by the blue arrows). Table 1 summarizes and Fig. 8 illus-
trates the results of this evaluation for all altitude levels, in-
dicating the number of available independent measurements
of the different datasets, and the theoretically derived stan-
dard errors of the mean. The decrease in the standard error is
a measure of the bene t obtained by the trade-off. The calcu-
lations of the standard errors in the mean when adding more
years do not change the curves for the restricted and full lon-
gitude ranges signi cantly, indicating that interannual varia-
tions are a negligible source of error. This justi es comparing
SPURT with ACE-FTS data, even though the measurements
are from different years.
Since ACE-FTS will continue measuring the atmosphere,
we will have the possibility to include more measurements in
a future comparison which then will improve the error assess-
(a) (b)
Fig. 8. Same as in previous Fig. 7, but now for all levels and for
(a) O3, and (b) H2O. Darker lines are calculations for the CMAM
dataset restricted to a latitude range of 40◦ N 60 ◦ N and a longitude
range of 0◦ E 45 ◦ E, gray lines include all longitudes. Calculations
are offset by 150, 90, 30, −30, −90, and −150% for O3, and by
280, 180, 80, −20, −120, −220% for H2O, respectively. Different
shadings of the colours represent calculations for the SPURT pe-
riod (2002 2003), the ACE-FTS period (2004 2007), and all years
(2002 2007), but note that these lines are hardly distinguishable.
Black and gray asterisks denote the number of independent ACE-
FTS measurements (N/b) for the restricted and full longitude ranges.
Red triangles denote the number of independent SPURT measure-
ments. Dashed lines indicate relative differences of 5% and 15%
for O3 and H2O, respectively.
ment. Or expressed in a different way, one can determine for
how many years the ACE-FTS observations should be con-
tinued in order to reduce the uncertainty in the climatology
below a certain level. For example, in order to obtain un-
certainties of 5% for O3 and 15% for H2O, respectively, 4 8
additional years of operation of the ACE-FTS would be re-
quired. However, we need also to nd larger but still reliable
data sets for comparison, since the given number of SPURT
measurements ultimately limits the achievable improvement,
especially in the troposphere and for H2O. It follows from
this evaluation that any measurement errors derived in the
comparison with SPURT have to be regarded as upper limits.
We move on to explain our ndings in Figs. 7 and 8 in a
more theoretical way. The errors in the measurements de-
rived by the suggested method include the variance intro-
duced by geophysical variability σ 2g , and the variance of the
actual measurement error σ 2m. For the full longitude range
the total squared error (f ) is then written as
f =
σ 2g + σ 2m
(N/b) . (3)
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1491
Table 1. Number of available independent measurements and theoretically derived relative differences (see text for details) for the ACE-FTS
and SPURT data at different levels. All are for 40◦ N 60 ◦ N and DJF 2004 2007.
dz from tp −5.5 km −2.5 km −0.5 km 0.5 km 2.5 km 5.5 km
N/b 1rel N/b 1rel N/b 1rel N/b 1rel N/b 1rel N/b 1rel
(a) O3
SPURT 6 9% 15 11% 42 6% 109 3% 16 8% %
ACE (45◦) 1 38% 2 29% 2 29% 3 18% 3 16% 4 18%
ACE (360◦) 2 24% 12 15% 13 13% 23 7% 34 6% 33 7%
(b) H2O
SPURT 2 55% 15 26% 42 15% 94 7% 16 15% %
ACE (45◦) 1 133% 2 66% 2 67% 3 40% 3 33% 5 5%
ACE (360◦) 2 74% 12 43% 12 46% 23 28% 34 18% 33 3%
For the restricted longitude range, the squared error (r )
changes to
r =
ασ 2g + σ 2m
β(N/b) , (4)
where α is the ratio of the geophysical variabilities, and β the
ratio of the number of measurements between the restricted
and the full elds. Both α and β are less than unity for a re-
stricted range. The issue is whether an expanded range leads
to reduced errors. This is a trade-off between the effects of α
and β. We now can de ne a trade-off factor γ
γ = r
f
=
ασ 2g + σ 2m
β(σ 2g + σ 2m)
. (5)
If γ>1 then full sampling is preferable to restricted sam-
pling. With
µ = σ
2
m
σ 2g
, (6)
Eq. (5) yields
γ = α + µ
β(1 + µ). (7)
If µ1 (i.e. geophysical variability dominates over measure-
ment error), then γ>1 if α>β, i.e. if the geophysical vari-
ance increases less rapidly than the number of measurements
as the range of longitudes increases. But for µ=O(1), the
condition is less stringent. This is shown in Fig. 9, where we
see that, except for H2O in the case of small measurement
errors, this requirement is true and it is bene cial to include
all data. This result is also consistent with what we expect
from Fig. 8 and Table 1, where moving from the restricted to
full longitude range is more bene cial in the case of O 3 than
in the case of H2O.
Fig. 9. Trade-off factor γ as a function of number of longitudes
included in the climatologies of O3 (solid lines) and H2O (dashed
lines). For the calculation of the trade-off factor, CMAM data from
the altitude bin 1 km below the tropopause and over 5 years of
Northern Hemisphere winter data were used. Different colours in-
dicate calculations performed for different µ’s, which represent the
ratio between the variance of the measurement error and the vari-
ance introduced by geophysical variability.
4 Results and discussion
4.1 SPURT versus ACE-FTS tracer-tracer correlations
Figure 10 shows the CO-O3 and H2O-O3 correlations for
the ACE-FTS and the SPURT measurements for the lati-
tude range between 30◦ N and 90◦ N, full longitude range,
and different seasons. ACE-FTS data were accumulated
over the years 2004 2007. The SPURT data were obtained
during two campaigns in two different years (2002/2003).
The agreement between the correlations of the two data sets
is remarkable, despite the fact that the measurements were
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1492 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
Fig. 10. CO-O3 (upper panels) and H2O-O3 correlations (lower panels) for ACE-FTS satellite (black) and SPURT aircraft (red) data between
30◦ N and 90◦ N. From left to right: winter, spring, summer, and autumn measurements. Note that the two data sets were obtained during
different years.
carried out in different years. In general, the SPURT data
lie well within the range of the ACE-FTS data, and the lim-
ited spread (or scatter) of the ACE-FTS indicates a very good
precision and vertical resolution of the instrument. Some fea-
tures such as the high-heel shape found in the spring H 2O-
O3 correlation, or the relatively broad transition between the
troposphere and the stratosphere in the summer CO-O3 cor-
relation, are evident in both data sets. With the lower verti-
cal resolution of a satellite instrument, the transition between
the stratospheric and the tropospheric branch in the tracer-
tracer correlation is expected to be less sharp. Using ACE-
FTS data from a narrower latitude range (e.g. between 40◦ N
and 60◦ N) does not change the results of this evaluation in a
signi cant way, it just tends to decrease slightly the spread in
the ACE-FTS correlation. Note that the ceiling altitude of the
aircraft was 14 km, hence the largest O3 values seen in the
ACE-FTS were not sampled. On the other hand, the ACE-
FTS reaches on average only down to an altitude of 5 km
which might explain why the largest H2O values in the air-
craft measurements were never sampled by the satellite. The
aircraft measurements in autumn which lie outside the ex-
pected H2O-O3 correlation range stem from an extraordinary
troposphere-to-stratosphere exchange event associated with
overshooting deep convection (Hegglin et al., 2004). Such
events are mixed into the background within a couple of days
and therefore are not a characteristic of the climatology.
4.2 SPURT versus ACE-FTS vertical pro les
The validation of seasonal ACE-FTS CO, O3, and H2O data
using vertical pro les relative to the tropopause height is pre-
sented in Fig. 11a, c and Fig. 12a, c. In order to minimize
the uncertainties caused by a potential latitudinal dependence
of the mean tracer pro les, we use data between 40 ◦ N and
60◦ N only. Based on the evaluation in Sect. 3.3, we use
ACE-FTS data obtained within the full longitude range be-
tween 180◦ W and 180◦ E. In the case of H2O, we also pro-
vide the evaluation for the restricted longitude range between
45◦ W and 45◦ E for comparison. Summer H2O data are not
shown, since the dynamical range of the SPURT measure-
ments did not allow a meaningful comparison (see Fig. 10,
third panel of the H2O-O3 correlation). The left panels of
Figs. 11 and 12 show the mean vertical pro les and their
standard deviations for the SPURT (black) and the ACE-FTS
data (gray). The right panels show the relative differences
and their uncertainties.
The comparison between satellite and aircraft data yields
the best result for CO. Average relative differences of the
mean are around ±9% in the UT, and ±12% in the LS. This
result is comparable to the 16% relative difference derived
from a validation using MOZAIC aircraft data in the UTLS
by Clerbaux et al. (2007). For O3, the derived relative dif-
ferences in the LS are ±19% on average, but closer to the
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1493
(a) (b) (c) (d)
Fig. 11. Comparison of ACE-FTS and SPURT CO and O3 measurements. Left panels in (a, c): Vertical pro les of mean CO and O 3
mixing ratios with standard deviations (1σ ) as a function of altitude relative to the thermal tropopause for SPURT (black) and ACE-FTS
(gray), and for DJF, MAM, JJA, and SON (from top to bottom), respectively. Left panels in (b, d): Same as in (a, c) but the SPURT vertical
CO and O3 pro les are smoothed according to Eq. (8) in order to account for the limited vertical resolution of the ACE-FTS instrument.
Right panels in (a, b, c, d): Relative differences between ACE-FTS and SPURT mean pro les (black line, calculated using Eq. (1) in text),
with positive/negative values indicating a high/low bias in the ACE-FTS measurements. The gray bar indicates ±10% relative difference.
Horizontal bars show the uncertainties in the relative differences. The total numbers of measurements are indicated for each altitude bin and
data set.
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1494 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
(a) (b) (c) (d)
Fig. 12. Same as previous Fig. 10 but for H2O. (a, b) ACE-FTS pro les used to calculate the mean pro les were taken between 45 ◦ W and
45◦ E, and (c, d) between 180◦ W and 180◦ E. From top to bottom: DJF, MAM, and SON. JJA comparison is omitted due to a too small
dynamical range of the SPURT H2O measurements.
5±5% derived by Dupuy et al. (2008) for altitudes higher
than 15 km if the autumn values between 3 and 6 km above
the tropopause are excluded from the calculation of the av-
erage. This is justi ed by the evaluation in Fig. 4d, which
suggests not including these altitude levels into a compari-
son. However, in the UT, the relative differences increase
to around 25% indicating a persistent high bias of the ACE-
FTS O3. This is consistent with a high bias in the ACE-FTS
O3 interpolated to the tropopause level with respect to MLS,
and other solar occultation instruments as shown by Manney
et al. (2007). The reason for this bias might be a sensitiv-
ity problem of the ACE-FTS when it comes to detecting and
accurately measuring low mixing ratios below atmospheric
layers with very high mixing ratios. Furthermore, at lower
altitudes, fewer microwindows are being used for the O3 re-
trieval and there are potentially more interfering species. In-
deed, the sensitivity problem is re ected in larger relative
retrieval errors in the troposphere. H2O shows the largest
relative differences of about ±18% in the LS, and ±30%
in the UT, with the uncertainties of the relative differences
suggesting a systematic low bias of the ACE-FTS measure-
ments, at least for the middle to upper troposphere in winter
and spring. However, H2O is the most variable tracer in the
troposphere among the species presented here. Indeed the
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1495
(a) (b) (c) (d)
Fig. 13. Comparison of ACE-FTS O3 and ozonesonde data. From top to bottom: DJF, MAM, JJA, and DJF. Left panels: vertical mean
pro les ±1σ standard deviations for ACE-FTS (in gray) and the ozonesonde data (in black). Right panels: Relative differences between the
two data sets and their uncertainties. The total numbers of measurements are indicated for each altitude bin and data set. Pro les were taken
(a) between 45◦ W and 45◦ E and during the years 2004 2007, (b) between 180◦ W and 180◦ E and during the years 2004 2007, (c) same
as in (b) but during the year 2005, (d) same as in (b) but during the year 2006.
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1496 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
Fig. 14. Same as previous Fig. 13b but with ozonesonde data smoothed according to Eq. (8).
evaluation in Sect. 3.3 indicates that up to 20% of these er-
rors might still be due to undersampling despite the improve-
ment when moving from the restricted to the full longitude
range. Including future measurements of the ACE-FTS into
this comparison will likely help to decrease the errors further,
however, the number of available SPURT measurements ulti-
mately will set another limitation. Hence we will need to nd
other reliable reference data sets for further comparisons.
While the vertical resolution of the ACE-FTS is affected
by its eld-of-view, as shown in Fig. 3 there is often a signif-
icant altitude oversampling in the vicinity of the tropopause,
which may increase the effective vertical resolution of the
measurements. This is also suggested by Fig. 10. If the reso-
lution was not better than 3 km, then the relatively sharp tran-
sition between the tropospheric and the stratospheric branch
of the correlation would be smeared out to a much greater
extent. However, at the level of 1 km, there is additionally
an inherent smoothing in the retrieval process, since forward
model calculations employ data on a 1-km grid. To account
(very roughly) for the limitations in the ACE-FTS vertical
resolution, we smoothed the SPURT pro les with a triangu-
lar function. We apply the following smoothing
χ [i]=0.25×χ [i−1]+0.5×χ [i]+0.25×χ [i+1] (8)
where i is the altitude level of a given tracer volume mixing
ratio (χ ). By smoothing the SPURT O3, CO, and H2O pro-
les with the above equation, we can gauge whether smear-
ing effects from the ACE-FTS vertical resolution have a sig-
ni cant effect on the results.
Figures 11b, d and 12b, d show the relative difference be-
tween the smoothed SPURT and the ACE-FTS mean tracer
pro les. The comparison with the unsmoothed vertical mean
pro les suggests that the smoothing effect accounts only for
about 5 15% of the total measurement error, with largest im-
pact just above and below the tropopause. Indeed, largest
impact is expected to be found close to the tropopause where
strong gradients in the tracer mixing ratios are present. Ac-
cordingly, CO seems to be the tracer which is least impacted
by the smoothing effect since it exhibits the weakest gradient
in tracer mixing ratios across the tropopause. This evaluation
suggests that the ACE-FTS is capable of resolving the transi-
tion between the troposphere and the stratosphere with high
accuracy, and points towards a vertical resolution of around
1 km.
4.3 Ozonesonde versus ACE-FTS vertical pro les
In order to gain more con dence in the validation results ob-
tained in the comparison with aircraft data, which may suffer
from the limited number of independent measurements, we
also use ozonesonde data which according to our evaluation
in Sect. 3.3 represent a perfect data set for a non-coincident
comparison. We look at different time and spatial samples in
order to explore, in an empirical way, how sampling issues
might in uence the validation results, and if these results are
consistent with what we expect from theory (Sect. 3.3).
In Fig. 13a, we show the seasonal comparison between the
ozonesonde and the ACE-FTS O3 data obtained during the
full observation period 2004 2007 and restricted to a longi-
tude band between 45◦ W and 45◦ E. In general, the results
obtained are comparable to the results found from the vali-
dation using the SPURT data set. The data show good agree-
ment in the LS, and an ACE-FTS high bias in the UT. How-
ever, the noise between the altitude levels is largely reduced,
related to the increase in number of measurements when us-
ing ozonesonde data instead of SPURT data. In Fig. 13b, the
comparison is shown between the data of the two data sets
taken within the full longitude range. This increases the num-
ber of measurements to compare with by a factor of 2 or 3 for
the ozonesondes, and by a factor of 4 for ACE-FTS. Includ-
ing more measurements leads to a reduction in the relative
differences but also in their uncertainties, hence this compar-
ison yields a statistically more meaningful error assessment
than using the restricted longitude range. The relative differ-
ences between the ACE-FTS O3 and ozonesonde measure-
ments are ±18% and ±8% for the UT and LS, respectively.
The ±8% in the LS is consistent with the results of Dupuy
et al. (2008). Finally, we compare the data obtained during
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1497
the years 2005 and 2006 separately (Fig. 13c and d, respec-
tively). The derived relative differences of the measurements
remain basically unchanged. A change can however be seen
in the uncertainties in the relative differences, which increase
due to the decrease in the available number of measurements.
This is consistent with our earlier analysis based on CMAM
and shown in Fig. 8, suggesting that inclusion of ozone mea-
surements taken during different years does not add a dis-
cernible amount of geophysical variability.
The validation using SPURT aircraft and ozonesonde data
thus yields consistent results, and the uncertainties are seen
to follow the theoretical expectations derived in Sect. 3.3. In
particular, the measurement errors are better characterized
when including measurements obtained over larger spatial
and temporal scales, despite the increase in geophysical vari-
ability. From this it follows that the trade-off factor (γ ) is
also >1 in the real atmosphere.
In Fig. 14, we nally show for each season the compari-
son between the ACE-FTS O3 data and the ozonesonde pro-
les smoothed according to Eq. (8) using all available years
and the full range of longitudes. The smoothing effect leads
to a decrease in the relative differences which is again most
pronounced just below the tropopause, and seen to be 5%
larger than in the comparison with the SPURT data set. The
larger decrease might re ect the additional bias introduced
by calculating tropopauses for the ozonesondes from simul-
taneously measured temperature pro les instead of from the
GEOS-4 analyses.
5 Conclusions
In this study we present the validation of ACE-FTS CO, O3,
and H2O measurements from Canada’s SCISAT-1 satellite in
the UTLS using SPURT aircraft and ozonesonde measure-
ments. This contributes to the validation efforts of the ACE-
FTS data published in this special issue. Other studies (ex-
cept the one focusing on CO), however, mainly focus on val-
idation in the stratosphere and the mesosphere. In the UTLS,
validation of chemical trace gas measurements is a challeng-
ing task due to small-scale variability in the tracer elds,
strong gradients of the tracers across the tropopause, and
scarcity of measurements suitable for validation purposes.
We here suggest two alternative/complementary methods
for the validation of satellite measurements in the UTLS
which are increasingly being used for the validation of chem-
istry transport models or chemistry climate models: tracer-
tracer correlations, and vertical pro les relative to tropopause
height. These methods are known to reduce geophysical vari-
ability, and thereby provide an instantaneous climatology ,
which avoids the need for coincident measurements. The cli-
matological comparison allows one to include all available
measurements, not just coincident ones, and thus obtain bet-
ter statistics and more reliable information about the instru-
ment accuracy and precision.
We rst assessed the performance and limitations of the
new validation methods. The most important question is
whether spatial and temporal variations of the climatology
are smaller than the errors one is trying to identify. This is
quanti ed through a trade-off factor. Using CMAM data to
simulate real measurements, we show that the uncertainty
in climatologies due to limited sampling is small enough
to allow for a meaningful error assessment in the UTLS if
enough measurements are available to describe the clima-
tology. However, this is not always the case, especially for
lower altitude levels. For the ACE-FTS, we show that in-
cluding more measurements obtained over a wider longi-
tude range improves the error assessment, despite the in-
creased geophysical variability. Given the number of cur-
rently available measurements, the uncertainty of the ACE-
FTS and SPURT climatologies is estimated to be around 5
10% for O3, and 10 25% for H 2O depending on the altitude
level. In contrast, standard validation techniques indicate that
ACE-FTS measurement errors in the UTLS region (i.e. be-
low 100 hPa or 15 km) may be as large as 50% or not deter-
minable (Dupuy et al., 2008; Carleer et al., 2008). Hence the
error assessment of the ACE-FTS can be improved signi -
cantly compared to standard validation techniques using the
new methods.
The theoretical analysis also allows one to estimate the
number of measurements needed to reduce the uncertain-
ties in the climatologies below a certain level. For exam-
ple, to obtain uncertainties of ±5% for O3 and ±15% for
H2O, around 50 100 and 80 120 independent measurements
would be needed, respectively, depending on the altitude
level. These numbers would require an additional 4 8 years
of operation of the ACE-FTS instrument.
From the ozonesonde measurements we determined that
using tropopause coordinates is only bene cial close to the
tropopause, i.e. around 5(4) km above(below) the tropopause
in winter and spring, and around 3(2) km in summer and au-
tumn, respectively. On the other hand, the correlation method
can be used wherever correlations are compact, i.e. for trac-
ers whose lifetimes are longer than the transport timescales
involved. This is true for both O3-CO, and O3-H2O in the
UTLS. In the stratosphere, the method would be applicable
to correlations involving N2O, CH4, and NOy. Although not
shown in this study, it will also be possible to extract quan-
titative information about the accuracy of one tracer, as long
as the accuracy of the other tracer is known.
We then applied the new methods to the measurements.
We used the tracer-tracer correlations to obtain qualitative
information on the precision, and the vertical pro les relative
to the tropopause to determine quantitatively the uncertainty
of the ACE-FTS measurements.
Using the correlation method, we nd that the ACE-FTS
measurements reproduce characteristic features found in the
SPURT aircraft measurements, and hence offer unprece-
dented precision for a satellite instrument. The evaluation
also suggests a vertical resolution of around 1 km. This
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008

1498 M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS
implies that the resolution is not solely determined by the
eld-of-view, which is around 3 km for the ACE-FTS. It
seems more important that the ACE-FTS oversamples the at-
mosphere, i.e. that the vertical spacing between two retrieval
levels is found to be <1 km around tropopause altitudes.
Comparison of vertical pro les in tropopause coordinates
yields upper bounds of the relative differences in the mean
between the ACE-FTS and SPURT aircraft measurements in
the UT and LS. These are found to be ±9% and ±12% for
CO, ±30% and ±18% for H2O, and ±25% and ±19% for
O3, respectively. In the troposphere, H2O shows a persistent
low bias in winter and spring, and O3 a persistent high bias
throughout the year. Comparison with ozonesonde data fur-
ther improves the error assessment, as predicted by the theo-
retical derivation of the trade-off factor. Relative differences
now are found to be ±18% and ±8% for the UT and LS,
respectively, and also their uncertainties decrease relative to
the ones obtained in the SPURT evaluation.
When taking into account the smearing effect of the
limited spacing between the measurements of the ACE-
FTS instrument, the errors decrease by 5 15% around the
tropopause. This limitation in the vertical resolution must
be taken into account when interpreting the ACE-FTS UTLS
measurements. Note that the tropopause heights of the dif-
ferent data sets used in validation studies should preferably
be derived using the same algorithms and analyses (e.g.
GEOS-4, GEOS-5, MetO (Met Of ce), or ECMWF (Euro-
pean Centre for Medium-Range Weather Forecasts)), since
differences in their characteristics may present an additional
error source. The signi cance of the results presented here,
especially for the H2O evaluations, should be tested with fur-
ther aircraft, balloon, and satellite data. Nevertheless, the
analysis indicates that the ACE-FTS offers unprecedented
precision and vertical resolution in the UTLS for a satel-
lite instrument, allowing a new global perspective on UTLS
trace gas distributions. Moreover, our study shows that air-
craft observations provide valuable data sets for satellite val-
idation, especially for tracers which are not routinely being
measured in the atmosphere (e.g. CO). By applying the meth-
ods presented here, more value can be extracted from his-
torical data sets. It furthermore allows the comparison of
satellites whose observation periods do not overlap, an issue
frequently pointed out in the literature.
Acknowledgements. The Atmospheric Chemistry Experiment
(ACE), also known as SCISAT, is a Canadian-led mission mainly
supported by the Canadian Space Agency and the Natural Sciences
and Engineering Research Council of Canada. Work at the Jet
Propulsion Laboratory, California Institute of Technology was
done under contract with the National Aeronautics and Space
Administration. M. Hegglin has been supported by the Cana-
dian Foundation for Climate and Atmospheric Sciences and the
Canadian Space Agency through the C-SPARC program. Thanks
to the WOUDC for providing ozonesonde data. Special thanks
to E. Dupuy for assistance with the ozonesonde data, and to
C. McLandress for helpful comments on the manuscript.
Edited by: T. Wagner
References
Beagley, S. R., de Grandpr·e, J., Koshyk, J. N., McFarlane, N. A.,
and Shepherd, T. G., Radiative-dynamical climatology of the
rst-generation Canadian Middle Atmosphere Model, Atmos.-
Ocean, 35, 293 331, 1997.
Bernath, F., McElroy, C. T., Abrams, M. C., et al.: Atmospheric
Chemistry Experiment (ACE): Mission overview, Geophys. Res.
Lett., 32, L15S01, doi:10.1029/2005GL022386, 2005.
Bloom, S. C., da Silva, A. M., Dee, D. P., et al.: The Goddard Earth
Observing Data Assimilation System, GEOS DAS Version 4.0.3:
Documentation and Validation, NASA, 104606 V26, 2005.
Boone, C. D., Nassar, R., Walker, K. A., et al.: Retrievals for the At-
mospheric Chemistry Experiment Fourier Transform Spectrom-
eter, Appl. Opt., 44, 7218 7231, 2005.
Carleer, M. R., Boone, C. D., Walker, K. A., et al.: Validation of
water vapour pro les from the Atmospheric Chemistry Experi-
ment (ACE), Atmos. Chem. Phys. Discuss., 8, 4499 4559, 2008,
http://www.atmos-chem-phys-discuss.net/8/4499/2008/.
Clerbaux, C., Coheur, P.-F., Hurtmans, D., Barret, B., Carleer,
M., Colin, R., Semeniuk, K., McConnell, J. C., Boone, C.,
and Bernath, P.: Carbon monoxide distribution from the ACE-
FTS solar occultation measurements, Geophys. Res. Lett., 32,
L16S01, doi:10.1029/2005GL022394, 2005.
Clerbaux, C., George, M., Turquety, S., et al.: CO measurements
from the ACE-FTS mission: validation using ground-based, air-
borne and satellite observations, Atmos. Chem. Phys. Discuss.,
7, 15 277 15 340, 2007.
de Grandpr·e, J., Beagley, S. R., Fomichev, V. I., Grif oen, E., Mc-
Connell, J. C., Medvedev, A. S., and Shepherd, T. G., Ozone
climatology using interactive chemistry: Results from the Cana-
dian Middle Atmosphere Model, J. Geophys. Res., 105, 26 475
26 491, 2000.
Douglass, A. R., Schoeberl, M. R., Rood, R. B., and Pawson,
S.: Evaluation of transport in the lower tropical stratosphere
in a global chemistry and transport model, J. Geophys. Res.,
108(D9), 4259, doi:10.1029/2002JD002696, 2003.
Dupuy, E., Walker, K. A., Kar, J., et al.: Validation of ozone mea-
surements from the Atmospheric Chemistry Experiment (ACE),
Atmos. Chem. Phys. Discuss., 8, 2513 2656, 2008,
http://www.atmos-chem-phys-discuss.net/8/2513/2008/.
Engel, A., B¤onisch, H., Brunner, D., et al.: Highly resolved obser-
vations of trace gases in the lowermost stratosphere and upper
troposphere from the Spurt project: an overview, Atmos. Chem.
Phys., 6, 283 301, 2006,
http://www.atmos-chem-phys.net/6/283/2006/.
Eyring, V., Butchart, N., Waugh, D. W., et al.: Assessment of tem-
perature, trace species and ozone in chemistry-climate model
simulations of the recent past, J. Geophys. Res., 111, D22308,
doi:10.1029/2006JD007327, 2006.
Fischer, H., Wienhold, F. G., Hoor, P., Bujok, O., Schiller, C.,
Siegmund, P., Ambaum, M., Scheeren, H. A., and Lelieveld, J.:
Tracer correlations in the northern high latitude lowermost strato-
sphere: In uence of cross-tropopause mass exchange, Geophys.
Res. Lett., 27(1), 97 100, doi:10.1029/1999GL010879, 2000.
Froidevaux, L., Livesey, N. J., Read, W. G., et al.: Early validation
analyses of atmospheric pro les from EOS MLS on the Aura
Atmos. Chem. Phys., 8, 1483 1499, 2008 www.atmos-chem-phys.net/8/1483/2008/

M. I. Hegglin et al.: ACE-FTS satellite validation in the UTLS 1499
satellite, IEEE Transaction on Geoscience and Remote Sensing,
44(5), 1106 1121, 2006.
Froidevaux, L., Jiang, Y., Lambert, A., et al. : Validation of Aura
Microwave Limb Sounder stratospheric ozone measurements, J.
Geophys. Res., doi:10.1029/2007JD008771, in press, 2008.
Fussen, D., Vanhellemont, F., Dodion, J., Bingen, C., Walker, K. A.,
Boone, C. D., McLeod, S. D. and Bernath, P. F.: Initial intercom-
parison of ozone and nitrogen dioxide number density pro les re-
trieved by the ACE-FTS and GOMOS occultation experiments,
Geophys. Res. Lett., 32, L16S02, doi:10.1029/2005GL022468,
2005.
Hegglin, M. I., Brunner, D., Wernli, H., et al.: Tracing troposphere-
to-stratosphere transport above a mid-latitude deep convective
system, Atmos. Chem. Phys., 4, 741 756, 2004,
http://www.atmos-chem-phys.net/4/741/2004/.
Hegglin, M. I., Brunner, D., Peter, T., et al.: Measurements of NO,
NOy, N2O, and O3 during SPURT: implications for transport and
chemistry in the lowermost stratosphere, Atmos. Chem. Phys., 6,
1331 1350, 2006,
http://www.atmos-chem-phys.net/6/1331/2006/.
Hegglin, M. I., and Shepherd, T. G.: O3-N2O correlations from
the Atmospheric Chemistry Experiment: Revisiting a diagnostic
of transport and chemistry in the stratosphere, J. Geophys. Res.,
112, D19301, doi:10.1029/2006JD008281, 2007.
Hoor, P., Gurk, C., Brunner, D., Hegglin, M. I., Wernli, H., and
Fischer, H.: Seasonality and extent of extratropcial TST derived
from in-situ CO measurements during SPURT, Atmos. Chem.
Phys., 4, 1427 1442, 2004,
http://www.atmos-chem-phys.net/4/1427/2004/.
Jin, J. J., Semeniuk, K., Jonsson, A. I., et al.: Co-located ACE-
FTS and Odin/SMR stratospheric-mesospheric CO 2004 mea-
surements and comparison with a GCM, Geophys. Res. Lett., 32,
L15S03, doi:10.1029/2005GL022433, 2005.
Krebsbach, M., Brunner, D., G¤unther, G., Hegglin, M., Maser, R.,
Mottaghy, D., Riese, M., Spelten, N., Wernli, H., and Schiller, C.:
Seasonal cycles and variability of O3 and H2O in the UT/LMS
during SPURT, Atmos. Chem. Phys., 6, 109 125, 2006,
http://www.atmos-chem-phys.net/6/109/2006/.
Manney, G. L., Daffer, W. H., Zawodny, J. M., et al.: Solar occulta-
tion satellite data and derived meteorological products: Sampling
issues and comparisons with Aura Microwave Limb Sounder,
J. Geophys. Res., 112, D24S50, doi:10.1029/2007JD008709,
2007.
Monahan, K. P., Pan, L. L., McDonald, A. J., Bodeker, G. E.,
Wei, J., George, S. E., Barnet, C. D., and Maddy, E.: Valida-
tion of AIRS v4 ozone in the UTLS using ozonesondes from
Lauder, NZ and Bolder, USA, J. Geophys. Res., 112, D17304,
doi:10.1029/2006JD008181, 2007.
McHugh, M., Magill, B., Walker, K. A., Boone, C. D., Bernath, P.
F., and Russell III, J. M.: Comparison of atmospheric retrievals
from ACE and HALOE, Geophys. Res. Lett., 32, L15S10,
doi:10.1029/2005GL022403, 2005.
Pan, L. L., Randel, W. J., Gary, B. L., Mahoney, M. J., and
Hintsa, E. J.: De nitions and sharpness of the extratropical
tropopause: A trace gas perspective, J. Geophys. Res., 109,
D23103, doi:10.1029/2004JD004982, 2004.
Pan, L. L., Wei, J. C., Kinnison, D. E., Garcia, R. R., Wuebbles,
D. J., and Brasseur, G. P.: A set of diagnostics for evaluating
chemistry-climate models in the extratropical tropopause region,
J. Geophys. Res., 112, D09316, doi:10.1029/2006JD007792,
2007.
Plumb, R. A. and Ko, M. K. W.: Interrelationships between mixing
ratios of long-lived stratospheric constituents, J. Geophys. Res.,
97, 10 145 10 156, 1992.
Reichler, T., Dameris, M., and Sausen, R.: Determining the
tropopause height from gridded data, Geophys. Res. Lett.,
30(20), 2042, doi:10.1029/2003GL018240, 2003.
Sankey, D. and Shepherd, T. G.: Correlations of long-lived chem-
ical species in a middle atmosphere general circulation model,
J. Geophys. Res., 108(D16), 4494, doi:10.1029/2002JD002799,
2003.
Strahan, S. E., Duncan, B. N., and Hoor, P.: Observationally de-
rived transport diagnostics for the lowermost stratosphere and
their application to the GMI chemistry and transport model, At-
mos. Chem. Phys., 7, 2435 2445, 2007,
http://www.atmos-chem-phys.net/7/2435/2007/.
Stratospheric Processes and their Role in Climate (SPARC): As-
sessment of trends in the vertical distribution of ozone, SPARC
report No.1, WMO Global Ozone Res. and Monit. Project Rep.
43, Editors N. Harris, R. Hudson and C. Phillips, World Meteo-
rol. Org., Geneva, 1998.
WMO: Meteorology A three-dimensional science, WMO Bul-
letin, 6, 134 138, 1957.
Walker, K. A., Randall, C. E., Trepte, C. R., Boone, C. D., and
Bernath, P. F.: Initial validation comparisons for the Atmo-
spheric Chemistry Experiment (ACE-FTS), Geophys. Res. Lett.,
32, L16S04, doi:10.1029/2005GL022388, 2005.
www.atmos-chem-phys.net/8/1483/2008/ Atmos. Chem. Phys., 8, 1483 1499, 2008