1

Comparison of POAM III ozone measurements with correlative
aircraft and balloon data during SOLVE

Jerry D. Lumpe & Mike Fromm
Computational Physics, Inc., Springfield, VA.

Karl Hoppel & Richard M. Bevilacqua
Naval Research Laboratory, Washington DC.

Cora E. Randall
Laboratory for Atmospheric & Space Physics, University of Colorado, Boulder, CO

Edward V. Browell & William B. Grant
Atmospheric Sciences Research, NASA Langley Research Center, Hampton, VA

Thomas McGee & John Burris
NASA Goddard Space Flight Center, Greenbelt, MD

Laurence Twigg
Science Systems and Applications, Inc, Lanham, MD

Erik C. Richard
Aeronomy Laboratory, National Oceanic and Atmospheric Administration,
Boulder, CO

Geoffrey C. Toon, James J. Margitan, & Bhaswar Sen
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA

Klaus Pfeilsticker, Hartmut Boesch, & Richard Fitzenberger
Institut fuer Umweltphysik, University Heidelberg, Heidelberg, Germany

Florence Goutail & Jean-Pierre Pommereau
Service d'Aeronomie du CNRS
BP3, 91371 VERRIERES LE BUISSON, France

2

Abstract

The Polar Ozone and Aerosol Measurement (POAM) III instrument operated
continuously during the SOLVE mission, making approximately 1400 ozone profile
measurements at high-latitudes both inside and outside the Arctic polar vortex. The
wealth of ozone measurements obtained from a variety of instruments and platforms
during SOLVE provided a unique opportunity to compare correlative measurements with
the POAM III dataset. In this paper we validate the POAM III version 3.0 ozone against
measurements from seven different instruments that operated as part of the combined
SOLVE/THESEO 2000 campaign. These include the airborne UV Differential
Absorption Lidar (UV DIAL) and the Airborne Raman Ozone and Temperature Lidar
(AROTEL) instruments on the DC-8, the dual-beam UV-Absorption Ozone Photometer
on the ER-2, the MkIV Interferometer balloon instrument, the Laboratoire de Physique
Molèculaire et Applications and Differential Optical Absorption Spectroscopy
(LPMA/DOAS) balloon gondola, the JPL in situ Ozone instrument on the OMS balloon
platform, and the Système D'Analyse par Observations Zénithales (SAOZ) balloon
sonde. The resulting comparisons show a remarkable degree of consistency despite the
very different measurement techniques inherent in the datasets, and thus provide a strong
validation of the POAM III version 3.0 ozone. This is particularly true in the primary 14
to 30 km region, where there are significant overlaps with all seven instruments. At these
altitudes POAM III agrees with all the datasets to within 7-10 % with no detectable bias.
The observed differences are within the combined errors of POAM III and the correlative
measurements. Above 30 km, only a handful of SOLVE correlative measurements exist
and the comparisons are highly variable, and therefore the results are inconclusive.
Below 14 km the SOLVE comparisons also show a large amount of scatter and it is

3
difficult to evaluate their consistency, although the number of correlative measurements
is large. The UV DIAL, DOAS, and JPL/OMS comparisons show differences of up to 15
% but no consistent bias. The ER-2, MkIV and SAOZ comparisons, on the other hand,
indicate a high POAM bias of 10 - 20 % at the lower altitudes. In general the SOLVE
validation results presented here are consistent with the validation of the POAM III
version 3.0 ozone using Stratospheric Aerosol and Gas Experiment (SAGE) II and
Halogen Occultation Experiment (HALOE) satellite data, and in situ ECC ozonesonde
data.

4

1.

Introduction

The objectives of the SAGE III Ozone Loss and Validation Experiment (SOLVE)
campaign were to examine the processes controlling ozone at mid- to high-latitudes and
to acquire multiple correlative data sets for validation of the Stratospheric Aerosol and
Gas Experiment (SAGE) III instrument. Successful completion of this objective required
a coordinated campaign of measurements in the Arctic high-latitude region using a
variety of instruments and platforms. These included the NASA DC-8 and ER-2 aircraft,
as well as balloon platforms, ground-based instruments and satellites. Of the latter the
Polar Ozone and Aerosol Measurement (POAM) III instrument was the only source of
high-latitude satellite vertical profile measurements used in the SOLVE campaign.

The initial motivation for introducing POAM III as a component of SOLVE was
primarily because its latitude coverage and measurement suite make it an ideal validation
platform for SAGE III. Figure 1 shows the POAM III latitude coverage in the Northern
Hemisphere compared to the predicted SAGE III ephemeris. The two instruments’
coverage overlaps a number of times during the year, providing multiple opportunities for
comparison. Furthermore POAM III measures the same primary species as SAGE III –
ozone, water vapor, nitrogen dioxide and aerosol extinction. Both POAM III and its
predecessor POAM II have used SAGE II extensively in their satellite validation studies
[Rusch et al., 1997, 2001; Randall et al., 2000, 2001a] and it is expected that both POAM
III and SAGE III will eventually benefit from a detailed retrieval comparison.

Of course, due to unforeseen launch delays, the SAGE III instrument was not
operational during the SOLVE campaign. Nevertheless, the wealth of high-latitude

5
measurements made during SOLVE provide a valuable resource for validation of the
POAM III dataset. In this paper we present validation of the POAM III version 3.0 ozone
using a number of coincident ozone measurements obtained during SOLVE. This study
complements POAM III ozone validation efforts which concentrate on comparisons with
other satellite data sets and ECC ozonesondes [Rusch et al., 2001]. The POAM
III/SOLVE ozone validation presented here includes aircraft and balloon data (both in
situ and remotely sensed).

This paper is not intended as a comprehensive intercomparison of the various
SOLVE ozone data sets but is presented strictly from the point of view of POAM III
validation. In addition to further quantifying the quality and scientific validity of the
POAM III ozone data, it is hoped that this study will prove useful in planning future
validation efforts for SAGE III. Absent an actual POAM III/SAGE III validation study,
which will have to await the SAGE III launch in 2001, we feel this is one of the most
useful contributions to SAGE III resulting from the POAM III involvement in SOLVE.
Of course, by helping to validate the POAM III data products, this effort will directly
benefit the SAGE III validation in the future.

In this work, the POAM III ozone measurements are compared with
measurements from seven different instruments that operated as part of the combined
SOLVE/THESEO 2000 campaign. These include the airborne UV Differential
Absorption Lidar (UV DIAL) and the Airborne Raman Ozone and Temperature Lidar
(AROTEL) instruments on the DC-8, the dual-beam UV-Absorption Ozone Photometer
on the ER-2, the MkIV balloon interferometer, the Laboratoire de Physique Molèculaire
et Applications and Differential Optical Absorption Spectroscopy (LPMA/DOAS)
balloon gondola, the JPL in situ Ozone instrument on the Observations of the Middle

6
Stratosphere (OMS) balloon platform, and the Système D'Analyse par Observations
Zénithales (SAOZ) balloon sonde. We first quantify the agreement between POAM III
and each of these instruments separately and then compare the results to see if they give a
consistent picture of the POAM III ozone validation, as well as maintaining consistency
with the satellite and ECC ozonesonde validation studies. We begin with an overview of
the POAM III ozone measurements and current status of the version 3.0 validation in
Section 2. In section 3 we compare the DC-8 data sets (UV DIAL and AROTEL),
followed by the ER-2 comparisons in section 4, MkIV in section 5, DOAS in section 6,
OMS/JPL Ozone in section 7, and finally the SAOZ comparisons in section 8. Section 9
contains a summary and conclusions.

2. Overview of POAM III measurements and version 3.0 ozone validation

The POAM III instrument is a nine-channel photometer that employs the
technique of solar occultation to derive composition and temperature throughout the
stratosphere and upper troposphere. By measuring atmospheric extinction in select bands
from 0.354 to 1.018 µm it is possible to retrieve density profiles of ozone, nitrogen
dioxide and water vapor, as well as temperature and wavelength-dependent aerosol
extinction. The instrument and its basic operational characteristics are described in detail
in Lucke et al. [1999].

POAM III has been in routine operation on the SPOT 4 satellite since April 24,
1998. It makes 14 measurements per day in each hemisphere, at approximately constant
latitude but separated in longitude by 25 degrees. This relatively coarse horizontal

7
sampling pattern is a consequence of the solar occultation geometry. Northern
Hemisphere (NH) measurements are made at satellite sunrise but actually correspond to
local sunset due to the retrograde orbit of the satellite. During the SOLVE time period
(November 1999 to March 2000) approximately 1400 NH measurements were made at
latitudes between 63.5 and 69°N.

Ozone is retrieved operationally between 60 km and a lower limit which is
typically in the mid- to upper-troposphere, depending on local cloud top height and
atmospheric opacity, which determines the minimum altitude to which the Sun sensor can
actively track the Sun. The POAM III version 3.0 retrieval algorithms and error analysis
are described in Lumpe et al. [2001]. The primary ozone information in the
measurements comes from the 603-nm channel, at the peak of the O3 Chappuis bands.
Based on the analysis presented in Lumpe et al. [2001] the total random error (precision)
of the POAM III ozone retrievals is estimated to be 3-5 % between 12 and 60 km,
increasing to 15 % or more at and below 10 km. In the lowermost stratosphere and upper
troposphere the ozone retrieval becomes very sensitive to accurate removal of the aerosol
extinction component. Maximum systematic errors due to cross section uncertainties are
estimated to be at the 1 to 2 % level.
The vertical resolution of the ozone retrieval, as defined by the width of the
retrieval averaging kernels, is 1 km throughout the stratosphere but degrades rather
quickly to 2-3 km in the upper troposphere [Lumpe et al. 2001]. Horizontal resolution
perpendicular to the instrument line of sight (i.e., parallel to the terminator) is limited by
the size of the solar disk, which is approximately 30 km at the tangent point. Parallel to
the line of sight one measure of horizontal resolution can be taken to be the path length of

8
the 1-km vertical shell sampling, which is approximately 200 km. However, this number
tends to underestimate the effective horizontal resolution since the information content in
the slant path measurement is sharply peaked at the tangent point.

This paper focuses on validation of the POAM III version 3.0 ozone using
correlative data obtained during the SOLVE/THESEO 2000 campaign. Preliminary
validation of an earlier version of POAM III ozone was presented in Lucke et al. [1999].
The version 3.0 ozone has also been validated against version 6.0 Stratosphere Aerosol
and Gas Experiment (SAGE) II and version 19 Halogen Occultation Experiment
(HALOE) satellite data, and balloon-borne ECC ozonesondes in Rusch et al. [2001]. The
results of this analysis show that the POAM III NH ozone agrees with SAGE II, HALOE
and ECC ozonesondes to within 5 to 7 %, with no bias, in the altitude range from 12 to
50 km. Between 50 and 60 km POAM tends to be biased high with respect to HALOE
by 5 – 10 %, but this bias is not seen in the POAM/SAGE II comparisons. Somewhat
larger disagreements (15 – 20 %) are seen below 12 km, with POAM III generally biased
high relative to both satellites and ECC ozonesondes. Unless otherwise noted, for the
remainder of this paper we will refer to POAM III simply as POAM to simplify the
notation.

3.

DC-8 comparisons.

3.1

Overview of DC-8 coincidences

9
The flight plans for the DC-8 were constructed specifically to provide a number
of direct coincidences with the POAM measurements. These overlaps provide an
excellent opportunity to compare measurements made by POAM and the DC-8
instruments in air masses that coincide closely in both time and space. There were six
such coincidences during SOLVE, occurring on December 2 & 14, January 16 & 25, and
March 3 & 9. On a number of these flights the DC-8 executed multiple passes through
the POAM tangent point, in addition to dives, to maximize the number of coincident
measurements made by both in situ and remote sensing instruments aboard the aircraft.

Figure 2 shows the location of the coincident measurements on each of these six
days. The red symbol represents the POAM 20-km tangent point and the red box
surrounding this point represents the area defined by the coincidence criteria used in the
comparisons. We define a coincidence as any measurements made within ± 1 deg in
latitude, ± 2 deg in longitude and ± 1 hour in time. These are significantly tighter criteria
than typically used in POAM satellite validation studies, but given the extent of the
overlaps they still yield a significant number of coincident measurements, as discussed
below. The blue line in Figure 2 represents the DC-8 flight track on each day. Although
difficult to see on the scale of this plot the flight path does pass through the coincidence
region on all days (more extensively on some days than others). Only those DC-8
measurements made within this red box (and satisfying the time constraint described
above) are used in the comparisons. For reference, contours corresponding to the
location of the outer, middle and inner edge of the polar vortex at 450 K have also been
plotted in the black solid and dashed lines. In Figure 2, and subsequently in this paper,
the convention of Nash et al. [1996] has been used to define the vortex edge.

10
These maps are helpful because it is important to take into account the position of
the measurements relative to the polar vortex in evaluating the ozone profile comparisons
below. For example, Figure 2 shows that on all six days except 12/2/99 the DC-8
traversed at least the inner edge of the vortex sometime during its flight. On three days
(12/2/99, 12/14/99 and 3/3/00) the POAM and DC-8 coincidences occur well inside the
vortex inner edge whereas on the other days the coincidences are near the vortex edge,
where one might expect strong horizontal gradients in the ozone. This will be discussed
in more detail below.

In this paper the POAM ozone measurements are compared with the two ozone
lidars operating on the DC-8; the NASA Langley UV Differential Absorption Lidar (UV
DIAL) and the NASA Goddard Airborne Raman Ozone and Temperature Lidar
(AROTEL). Both of these instruments measure the ozone profile above the aircraft and
therefore provide significant overlap with the POAM measurements in the low- to midstratosphere. There is an added advantage in being able to compare simultaneously with
two instruments that are essentially measuring the same air mass continuously. This is a
unique opportunity, which is further enhanced by the fact that all three instruments make
the same fundamental measurement – ozone concentration as a function of altitude.

3.2

Coincident data sets

The NASA Langley airborne UV Differential Absorption Lidar (UV DIAL)
system has been used to measure ozone, aerosol, and cloud profiles during four previous
stratospheric ozone investigations, affording the opportunity for many intercomparisons
with other ozone measuring instruments [Browell, 1989; Browell et al., 1990, 1993,

11
1998; Grant et al., 1998]. This system uses two frequency-doubled Nd:YAG lasers to
pump two high-conversion-efficiency, frequency-doubled, tunable dye lasers. In
stratospheric O3 investigations, the two frequency-doubled dye lasers are operated
independently with one tuned to 301 nm for the O3 DIAL on-line wavelength and the
other tuned to 311 nm for the off-line wavelength. All of the beams are transmitted in the
zenith direction through a 40-cm-diameter fused silica window. The atmospheric
backscattered laser returns are collected by a 36-cm diameter telescope, optically
separated, and directed on to different detectors.

The UV DIAL instrument measures ozone up to 10-15 km above the aircraft with
an accuracy of better than 10 %. Vertical resolution of the measurements is 750 m and
horizontal resolution is 70 km (5 min). Note that the horizontal and vertical resolutions
of the UV DIAL ozone measurements, although slightly better than POAM, are very
similar. The UV DIAL ozone has been compared with ECC ozonesonde data from Ny
Aalesund (79o N, 12 o E) during SOLVE [Grant et al., manuscript in preparation]. In the
15 – 20 km range the mean difference between the two measurements is ~ 3 % with a
standard deviation of 4 %. The sign of the difference is such that UV DIAL tends to
biased low relative to the ECC ozonesondes in this altitude range.

The Airborne Raman Ozone, Aerosol and Temperature Lidar (AROTEL) is a DC8 instrument conceived, designed and built by Goddard Space Flight Center in
collaboration with scientists from Langley Research Center. The instrument was flown
for the first time during the SOLVE campaign. The instrument is a multi-wavelength
lidar: radiation at 1064, 532 and 355 nm are transmitted from a Nd-YAG laser, and 308nm radiation is transmitted from a XeCl excimer laser. These wavelengths are transmitted

12
nearly simultaneously, and backscattered radiation is collected with a 16" telescope. This
returned light at the transmitted wavelengths, as well as Raman scattered radiation at 387
and 332 nm (N 2 scattering from the transmitted 355 and 308 nm beams) is wavelengthseparated using dichroic beamsplitters and detected using Hamamatsu R7400P
phototubes.

Ozone data is extracted from the four UV signals using the differential absorption
lidar technique. The algorithm for ozone retrieval is essentially identical to the
description in McGee et al. [1995], which describes the NDSC ground-based system. The
only difference is in the use of a fourth order polynomial fitting function as opposed to
the linear function previously described. In order to ensure linearity of signals, multiple
detectors are used for each wavelength. Beyond about 3 km above the aircraft the UV
signals are all photon counted; analog detection is used near to the aircraft. Vertical
resolution of the retrieved ozone profiles is approximately 1 km and the accuracy is better
than 10 %. A detailed description of the instrument is given in McGee et al. [2001].

3.3

Comparison of POAM, UV DIAL and AROTEL ozone

In this section we first discuss the comparisons between POAM and the two DC-8
instruments individually, and then look at how well the three data sets compare
collectively. For simplicity we have plotted the coincident ozone profiles from all three
instruments together in Figure 3. In each panel, corresponding to one DC-8 flight date,
the black curve represents the single ozone profile measured by POAM at the
coincidence location indicated by the red symbol in Figure 2. For the DC-8 instruments,
rather than plot all the individual profiles satisfying the coincidence criteria, we have

13
plotted a mean profile for each flight. For the UV DIAL data, plotted in red, this is
produced by binning all the coincident data (obtained directly from the SOLVE data
archive) in 0.25 km bins and then calculating an average ozone density in each bin. The
AROTEL raw data meeting the coincidence criteria were averaged prior to the retrieval
of ozone, instead of calculating the average of profiles within the archived data set. This
approach permitted the retrieval to reach to higher altitudes than the ozone profile
averaging method. The average AROTEL profiles are plotted in blue in Figure 3.

Note that the UV DIAL and AROTEL measurements tend to sample somewhat
different vertical regions of the stratosphere. Profiles from the two instruments generally
overlap in the altitude range from ~ 14 to 23 km. For the six days of interest in this study
ozone retrievals are typically available between about 11 to 23 km for UV DIAL and 14
to 30 km for AROTEL. Exceptions for AROTEL include January 25, where the data
below 19 have been removed due to possible PSC contamination and the December 14
and March 9 profiles, which extend all the way to 40 km, but with sharply increasing
random error above 30 km (not shown).

In general the POAM and UV DIAL ozone density profiles agree very well. On
most flights the vertical structure of the profiles measured by the two instruments are
very consistent and the magnitude generally agrees well. Notable exceptions are the low
altitude discrepancies on 1/16/00 and 3/3/00 and the fairly bad overall agreement on
3/9/00. The only systematic differences apparent in these comparisons is that POAM
tends to be high relative to UV DIAL above about 19 km. It is likely that this is due to
the tendency for the UV DIAL measurements to be slightly underestimated at the higher
altitudes in sunlit conditions. This tendency has been seen previously in comparisons

14
between UV DIAL and correlative SAGE II and HALOE measurements [Grant et al.,
1998].

The low altitude discrepancies on 1/16/00 and 3/3/00 are likely due to midlatitude intrusions and large gradients in the ozone distribution below 13-14 km, which
are readily seen in the UV DIAL flight images on those days. On 3/9/00, where we see
the greatest differences in the 14-20 km range, the measurements were very near the
vortex edge, and again very large horizontal gradients in ozone are evident in the UV
DIAL ozone cross section. Also, on this particular day the DC-8 coincidences were
actually biased towards low latitudes relative to the POAM point, which accentuates the
sampling bias due to the strong ozone gradients. For these reasons we believe the
discrepancies seen on this day, and at the low altitudes on 1/16/00 and 3/3/00, are more
indicative of true atmospheric variability than any fundamental error in either
measurement.

For the most part Figure 3 also shows quite good agreement between the POAM
and AROTEL profiles. Both instruments consistently reproduce small-scale vertical
structure in the ozone profile, although the lidar measurements often show more detailed
vertical structure than the POAM profiles (see, e.g., the January 16 and 25 profiles).
There also appears to be a tendency on some days for POAM to be low relative to
AROTEL at the peak of the ozone profile. For the most part however, the agreement in
the profiles above 20 km is good, and even up to 40 km for the two AROTEL profiles
that extend to that altitude. One notable exception is on March 3, where the AROTEL
ozone appears to have a noticeably smaller scale height than POAM above the profile
peak. The March 9 POAM/AROTEL comparison, like the POAM/UV DIAL comparison

15
on that day, shows poor agreement overall below 25 km. It is interesting to note,
however, that the two lidar profiles themselves do not agree very well on this day,
showing very different vertical structure and a clear systematic bias. At the present time
this is not understood.

Figure 4 summarizes the ozone differences between POAM and the DC-8
instruments. For each flight the average UV DIAL and AROTEL ozone profiles plotted
in Figure 3 were first linearly interpolated to the standard POAM 1-km altitude grid. The
relative difference profile (in percent) was then calculated according to

∆ = 200 ×

POAM − Corr
POAM + Corr

where Corr represents a correlative measurement, in this case UV DIAL or AROTEL.
The POAM/UV DIAL and POAM/AROTEL differences are plotted in the top and
middle panels, respectively, of Figure 4. In each panel the colored curves (red for UV
DIAL, blue for AROTEL) correspond to the individual difference profiles for each flight
and the black profile is the mean of all six flights. Error bars correspond to the standard
error of the mean difference.

The POAM/UV DIAL mean difference is within 5 to 7 % between 14 and 20 km,
and less than 10 % between 13 and 21 km. At the lowest altitude point, 12 km, the mean
error is clearly dominated by the large differences on 1/16/00 and 3/3/00 discussed above.
Also, the general tendency for UV DIAL to be biased somewhat low relative to POAM
above 20 km is evident in the mean. The fact that the UV DIAL ozone is biased low on

(1)

16
average by a few percent from 11 to 20 km is consistent with the UV DIAL/ECC
comparisons [Grant et al., manuscript in preparation], and probably indicates a slight
overestimate in the ozone absorption coefficient used in the DIAL analysis (note,
however, this should not affect the trend determination from UV DIAL data). Except for
these small systematics, which we feel are understood, the two instruments agree to
within the combined measurement uncertainties and these results are generally consistent
with previous POAM validation [Rusch et al., 2001].

The mean POAM/AROTEL difference is within 10 % at all altitudes below 37
km. The exception is the localized peak at 28-29 km where differences reach 10 - 20 %,
with POAM high (note, however, that this feature is dominated by the December 12
event where the AROTEL profile shows a pronounced minimum at this altitude which is
not seen by POAM). The results show a fairly consistent vertical structure in the
difference profiles for all days, with differences generally changing from -10 % at 18 km
to +10 % at 28 km. Note that above 30 km only two measurements, from the December
14 and March 9 flights, contribute to the mean. While these also appear to be fairly
consistent, given the small number of measurements in this altitude range it is difficult to
tell how significant this result is. The POAM/AROTEL differences in the 15 to 27 km
range are generally consistent with the POAM satellite and ECC ozonesonde validation,
although the peak difference of -10 % at 18-19 km (again a fairly systematic feature in
the six flights) is larger than the maximum differences seen in Rusch et al. [2001].

In summary, the bottom panel of Figure 4 shows the mean difference profiles for
the POAM/UV DIAL and POAM/AROTEL comparisons on the same plot. Again,
ignoring the POAM/UV DIAL differences above 20 km, POAM agrees with both

17
instruments to within 10 % between 13 and 27 km with larger errors (but oscillating
about zero) in the POAM/AROTEL comparisons at the higher altitudes. It is interesting
to note that there is a very similar systematic shape to the two mean difference profiles,
as well as a nearly constant offset. At this point these correlations in the differences
between the UV DIAL and AROTEL comparisons are not understood, but should be
explored more in the future.

4. ER-2 comparisons

During the two SOLVE deployments the ER-2 aircraft made a total of 11 science
flights out of Kiruna (neglecting transit flights). On each of those flights in situ
measurements of the ozone concentration were made from the Q-Bay of the aircraft by
the dual-beam UV-Absorption Ozone Photometer (hereafter referred to as simply
“NOAA Ozone”) [Proffitt et al., 1989]. The instrument consists of a 254-nm mercury
lamp, two sample chambers that can be periodically scrubbed of ozone, and two detectors
that measure the radiation directed from the lamp through the chambers. Ozone absorbs
strongly at this wavelength and the absorption cross-section is accurately known; hence,
the ozone number density can be accurately calculated from the difference in the detected
signals from the two chambers. Since the two absorption chambers are identical, virtually
continuous measurements of ozone are made by alternating the ambient air sample and
ozone scrubbed sample between the two chambers. At a one-second data collection rate,
the minimum detectable concentration of ozone (one standard deviation) is 1.5 x 1010
molecules/cm3 (0.6 ppbv at STP or 8 ppbv at 20 km). Measurement accuracy is predicted
to be 3% plus precision [Proffitt et al., 1989].

18
Because the NOAA Ozone measurements are made in situ and the aircraft tends
to maintain a more or less constant cruise altitude at approximately 18-20 km (aside from
takeoff, landing and occasional dives), the POAM/ER-2 comparisons are necessarily
heavily weighted to this narrow altitude range. In this region Rusch et al. [2001] has
shown that the POAM version 3.0 ozone agrees to within 3 - 5 % with coincident satellite
and ECC ozonesonde data in the Northern Hemisphere (with POAM in general somewhat
low compared to the satellites but high relative to the sondes).

Because of the problematic nature of comparing satellite to in situ measurements,
we have used three different approaches in comparing the POAM and ER-2 ozone data:
vortex-average, trajectory hunting, and direct coincidence measurements. In the first
approach, vortex-averaged ozone measurements from each instrument are compared for
all of the ER-2 flights. This approach was initially motivated by the observation that the
ER-2 ozone data from most of the deep vortex-survey flights generally shows a very
uniform ozone field within the interior of the vortex, at least in the early winter (see
discussion below). This uniformity is also apparent from the in-vortex measurements
made by POAM over time periods of several days. The second method used to compare
POAM and ER-2 ozone data involves using a trajectory analysis to identify and directly
compare identical air parcels that were sampled by both instruments. Finally, we have
also identified and compared standard temporal and spatial coincidences between the
POAM and ER-2 ozone measurements. Unlike the DC-8, no effort was made to
incorporate specific underflights of the POAM measurement locations into the ER-2
flight plans during SOLVE. However, direct coincidences do exist on a number of days
if the coincidence criteria are relaxed considerably from those used for the DC-8
comparisons.

19

Results from these three different comparison techniques are discussed separately
in sections 4.1- 4.3 below. All three techniques have their strengths and shortcomings.
The vortex-average method minimizes the effect of random errors, but depends upon
having unbiased (or at least the same) ER-2 and POAM vortex sampling. The trajectory
hunting method would, in principle, be the best way to ensure comparison of similar air
parcels, but it relies on the accuracy of the trajectory analysis. Finally, the direct
coincidence technique is the most straightforward and frequently used validation method,
but given the well-known trade-offs between constraining the closeness of the
coincidences and obtaining a statistically significant number of samples, there is no
guarantee that similar air parcels are sampled.

4.1

Vortex-average ozone comparisons

In order to compare the vortex-average ozone measured by POAM and the ER-2
it was first necessary to isolate the in-vortex data obtained by each instrument on a given
day. For the ER-2 the N2 O and CO2 tracer data were used to determine which time
segments of the flight corresponded to air sampled well within the interior of the vortex.
All data meeting the vortex-discrimination criteria was then binned in uniform 10 K
potential temperature bins to create a single vortex-average vertical profile.

As discussed in a companion paper in this special issue [Randall et al., 2001b],
POAM routinely made measurements both inside and outside the polar vortex on a daily
basis during SOLVE. This is because the center of the vortex was frequently displaced
from the pole towards Europe or Asia and therefore POAM, measuring around a circle of

20
latitude, would sample both in-vortex and extra-vortex air on a given day. However,
because of the relatively coarse horizontal sampling afforded by the solar occultation
technique, on any given day only a handful of events might be well inside the vortex.
Therefore, to obtain a statistically meaningful sample of in-vortex ozone measurements
for each ER-2 flight day the POAM data were averaged over a three-day time period
centered on the flight date (i.e., flight day ± 1 day). All POAM profiles measured within
the inner vortex boundary (as defined by the Nash criteria) in that three-day period were
averaged together and interpolated onto the same potential temperature grid as the ER-2
data.

Figure 5 illustrates this procedure for four representative ER-2 flight days. For
each day the ER-2 flight track is represented by the solid blue line, the inner and middle
Nash vortex edges (at 450 K) are represented by the black and green contours,
respectively, and all POAM measurements made during the three-day averaging period
are shown as red dots. Only those POAM points and that portion of the ER-2 flight lying
within the solid black contour contribute to the vortex-average profile (the POAM vortex
discrimination uses the appropriate vortex boundary for each of the three days used in the
average). The number of such points of course varies from flight to flight. On ER-2
vortex survey flights, such as 3/5/00, the entire ER-2 flight occurs deep in the vortex
whereas other flights contain planned vortex edge crossings.

Figure 6 shows the vortex-average ozone profiles measured by both instruments
for the eleven ER-2 science flights out of Kiruna. The red circles represent the POAM
points, black squares are the ER-2 points and the solid lines represent ± one standard
deviation of the mean for each flight. Clearly there is generally very good quantitative

21
agreement between the two data sets. POAM does appear biased somewhat low relative
to ER-2 on the January 27 and 31 flights, but by no more than 5 %. Figure 6 also shows
an increasing tendency for POAM ozone to be high relative to the ER-2 at the higher
altitudes in late winter, beginning with the March 5 flight. In this late-winter period the
vortex had experienced a large amount of chemical ozone loss [Hoppel et al., 2001], and
POAM measured large in-vortex ozone gradients [Randall et al., 2001b]. Figure 6 also
suggests that while the POAM vortex-average standard deviation is increasing over this
late winter period, the ER-2 variation is not. This suggests that the ER-2 was
preferentially sampling the most depleted vortex air, while POAM obtained more of a
vortex survey. This view is reinforced by Figure 5, which shows the small slice of the
vortex sampled by the ER-2 on March 11 relative to the more extensive POAM vortex
sampling (the corresponding March 12 map, which is not shown, yields the same
conclusion). Of course, it was not atypical for POAM to sample a larger portion of the
vortex than the ER-2 throughout the winter. However, increasing ozone gradients within
the vortex (as a result of ozone loss) later in the winter would serve to accentuate the
effects of this sampling bias. Thus, we suggest that the increasing tendency for POAM to
be biased high compared to the ER-2 in the late winter comparisons is much more likely
to be the result of sampling biases rather than real measurement differences.

To illustrate this point, in Figure 7 we have plotted the POAM and ER-2 ozone
measurements at 460 K as a function of equivalent latitude on January 23 and March 11
(equivalent latitude is the latitude enclosing the same area as a given potential vorticity
contour). The red symbols represent all the POAM measurements at 460 K within ± 1
day of flight date (both inside and outside the vortex) whereas black symbols represent
the ER-2 ozone data (in-vortex only). The vertical lines denote the middle and inner

22
vortex edges. On January 23 the ozone inside the vortex, as measured by both
instruments, is quite uniform. This situation is typical of the early- and mid-winter flights
and, as discussed previously, provided the original justification for using vortex-average
comparisons. Clearly the situation is very different by March 11. Here the POAM data
show evidence of large ozone gradients within the vortex, with two distinct populations
of low and high ozone; the ER-2, however, preferentially samples only that part of the
vortex with depleted ozone. This explains why the vortex-averaged POAM ozone on this
day, and on all other days in the late winter period, is biased high relative to the ER-2.
This is purely a sampling bias, rather than a measurement error.

Finally, we note that the sharp structures in the 1/14/00 and 1/23/00 ER-2 profiles
are due to intrusions (mid-latitude filaments) encountered by the ER-2 on those flights.
This can be clearly seen by correlating the O3 and tracer data from different ER-2
instruments on those flights (not shown). While POAM can in principle also see such
filaments, there is no guarantee that it sampled these specific features on those days.
Even if a filament were seen in one POAM measurement on the flight date it would be
smoothed out in the three-day averaging.

In Figure 8 these results are summarized by plotting the average ozone mixing
ratio profiles, and their mean difference, for the eleven flights. The mean difference is
within 5 % at all levels between 350 and 450 K and within 7 % at 460 K. The only
difference greater than 10 % is at the highest theta level, 470 K, which is clearly
dominated by sampling biases discussed above. On the whole, the vortex-averaged
comparisons show excellent agreement between POAM and ER-2 ozone.

23
4.2

Trajectory hunting analysis

The second method of comparing POAM and ER-2 measurements is to look for
air parcel trajectory matches. Forward and backward isentropic trajectories were run
from each POAM measurement location at 10 K increments in the vertical using the
UKMO wind analysis. We then searched for cases where these trajectories crossed the
ER-2 flight track within the following constraints: 300 km horizontal separation, 5 K
separation in the vertical, and 1.2 hours in time. The matches were further constrained to
have a maximum trajectory length of ± 5 days. Of course, it is possible that within five
days significant ozone chemical change could occur along the trajectory, but most of the
matches found in the analysis were within three days, and fairly evenly distributed
between forward and backward trajectories. A total of 249 matches were found to meet
the above criteria in the vertical range from 380 to 470 K.

The results of the trajectory hunting analysis are summarized in Figure 9. The
small black dots represent the ozone difference for all of the 249 points that satisfied the
match criteria. Note that there tends to be more scatter in trajectory-match comparisons
relative to the vortex-average comparisons. This is expected both because there is some
scatter introduced by trajectory inaccuracies and of course the vortex-average approach,
which averages over many measurements, naturally tends to produce a smoother result.
The black curve in Figure 9 represents the mean difference for all the match events (after
binning the data in 10 K potential temperature bins) and the error bars correspond to the
standard error of the mean difference. Comparing this result with Figure 8 shows that the
trajectory matching and vortex-average comparisons yield remarkably similar results in
terms of both the magnitude and vertical structure of the ozone difference. In particular,

24
it is interesting to note that the high POAM bias above 450 K appears in both analyses.
However, close inspection of the trajectory matching results shows that there is no trend
toward increasing POAM biases later in the winter as was observed in the vortex-average
comparisons. Furthermore, there is no observed bias with respect to forward or backward
trajectories, which could produce a bias in the relative differences if chemical depletion is
occurring along the trajectories. Therefore, the consistency of this high bias in the two
approaches is not currently understood and deserves further study.

4.3

Comparison of POAM/ER-2 direct coincidences

Finally, we present comparisons of POAM measurements made in direct temporal
and spatial coincidence with the ER-2 measurements. If the stringent coincidence criteria
used in the DC-8 analysis in section 3 are used for the ER-2, very few coincident
measurements are found. However, if the criteria are relaxed considerably many more
coincident data points are found. For the comparisons presented here we chose the
following criteria: ± 3 deg in latitude, ± 10 deg in longitude and ± 3 hour in time. These
criteria are similar to those typically used in the POAM satellite validation studies
[Randall et al., 2000, 2001a; Rusch et al., 1997, 2001]. We further required that both
POAM and ER-2 measurements be made within the inner edge of the vortex.

Coincident POAM and ER-2 measurements satisfying these criteria were found
for all the ER-2 flights discussed in section 4.1 except the flight on January 14. In
addition, all of these coincident ER-2 measurements contain some portion of the data
obtained during takeoff, landing or the executed dives and stack maneuvers. Therefore
they all contain ER-2 data over a range of altitudes, which allows for direct profile

25
comparisons with POAM. For these comparisons we plot the coincident POAM and ER2 data versus altitude since this is the fundamental vertical grid for the POAM data and is
also available directly for the ER-2 from the archived GPS altitudes.

Figure 10 shows the coincident POAM (in blue) and ER-2 (in red) ozone mixing
ratio profiles for each flight. The ER-2 profiles were constructed by binning all points
satisfying the coincidence criteria into 1-km altitude bins and calculating a mean ozone
mixing ratio in each bin. The resulting profile was then linearly interpolated to the
standard POAM altitude grid. The altitude range of the coincident ER-2 profiles varies
considerably from flight to flight but most flights provide a significant overlap with the
POAM profiles. In general the agreement is qualitatively quite good, although occasional
systematic differences do appear (e.g., January 23 below 16 km, February 2, and March
5). Note that the comparisons do not show the increasing tendency toward higher POAM
biases as was seen in the vortex-average and trajectory hunting comparisons. However,
there is little direct POAM/ER-2 overlap in the 460-470 K (~ 18 km) region where these
biases occur, particularly late in the winter.

Figure 11 shows the mean ozone difference profile (black curve) obtained from
averaging the coincident data from all ten flights. The numbers on the right axis
represent the number of profiles that contribute to the mean at each altitude level.
Between 11 and 19 km the differences are within 10 %. However there is a small but
statistically significant bias over most of this altitude range, with POAM higher than ER2 by an average of ~ 5 %. Below 11 km POAM is systematically higher than ER-2 by
approximately 20 % on average.

26
For comparison the mean difference profiles from the vortex-average and
trajectory matching analyses are included in the blue and red curves, respectively, in
Figure 11. The result of these three independent approaches lead to generally consistent
conclusions at the ± 5 % level, with mean difference profiles at most altitudes
overlapping within their standard errors. However the vortex-average and trajectory
matching results do not reproduce the POAM high bias between 12 and 17 km seen in the
standard coincidence comparisons. In fact, it could be argued that these two approaches
show a low POAM bias of a few percent, at least below 18 km. This discrepancy is not
currently understood. Finally, we note that these results are generally consistent with the
conclusions of Danilin et al. [2001], who have also done a systematic comparison of the
POAM and ER-2 ozone measurements during SOLVE using both a trajectory hunting
and direct coincidence analysis.

5.

MkIV balloon comparisons

The Jet Propulsion Laboratory MkIV Interferometer [Toon, 1991] is a highresolution Fourier transform infrared (FTIR) spectrometer. Like POAM, the MkIV
makes measurements of atmospheric composition using the solar occultation technique,
but operates in the infrared rather than the visible, and from a balloon rather than a
satellite platform. In terms of optical design the MkIV is very similar to the ATMOS
instrument [Farmer, 1987] which flew four times on the space shuttle. Following launch,
the balloon rises to float altitude (typically 30 to 40 km) from which it then observes the
Sun rise or set. When operating at high latitudes, such as during SOLVE, the sunrise/set
transition is relatively slow, taking approximately two hours. During occultation, the
instrument measures the solar radiance through the atmosphere in the spectral range from

27
650 to 5650 cm-1 (1.77 to 15.4 µm) at 0.01 cm-1 resolution. Normalization of the limb
spectra by a high-Sun (unattenuated) spectrum removes the solar and instrumental
features yielding atmospheric limb transmittance spectra. These transmittance spectra are
analyzed using a nonlinear least squares spectral fitting algorithm to determine slant
column abundances of various trace gases along the limb path from the depths of their
absorption lines. Finally, the matrix equation relating the measured slant column
abundances to the calculated slant path distances is solved to yield the final volume
mixing ratio profiles for each gas. The MkIV retrieval algorithms are described by Sen et
al. [1998]. The effective vertical resolution of the MkIV profiles is approximately 2 km.
The accuracy of the ozone retrieval, which is determined primarily by uncertainties in
spectroscopic parameters used in the line-by-line forward model, is estimated to be ~ 5
%.

The MkIV payload made two flights during SOLVE, launching from the Esrange
balloon facility just outside of Kiruna. On the first flight, December 3 1999, MkIV
measured a sunset occultation, whereas the second flight on March 15 2000 was a sunrise
event. During the time it takes for the MkIV to record an occultation, the balloon can
drift a significant distance horizontally, depending on the local wind fields. On each day
POAM made two measurements close in latitude to the balloon position and bracketing it
in longitude (see Figure 13 for details). The spatial separation between the balloon and
satellite measurements is similar on both days, but the different character of the balloon
track on the two flights introduces distinct differences. On the December flight the
balloon drifted in a predominantly north-south direction and the two POAM
measurements are almost equidistant in longitude from the MkIV and at the same average
latitude. On the March flight the balloon traveled almost entirely in an east-west

28
direction, bringing it in much closer spatial coincidence with one of the POAM
measurements on that day. The temporal coincidence was also very different on the two
flights. Because POAM measures local sunsets in the Northern Hemisphere, the two
instruments made measurements in close temporal coincidence on the first date (MkIV
sunset) but were separated by approximately 12 hours on the March flight (MkIV
sunrise).

Another important difference between the two flights is the location of the
measurements relative to the edge of the polar vortex. To look at this more closely we
have plotted in Figure 12 profiles of equivalent latitude for the POAM and MkIV
measurement locations, as well as the inner and middle vortex edges. The December 3
results in the left panel show POAM and MkIV sampling essentially identical equivalent
latitudes well inside the inner vortex edge. On March 15, shown in the right panel, the
situation is more complicated. The MkIV profile and the closest coincident POAM
profile, at 39 E longitude, are fairly close in equivalent latitude and both just inside the
vortex inner edge. The other POAM profile, at 14 E longitude, is clearly sampling
different equivalent latitudes, more consistent with the middle of the vortex boundary.

In Figure 13 the POAM and MkIV ozone profiles are plotted for the two days. In
both panels the black curve is the MkIV ozone and the blue and red curves correspond to
the two coincident POAM profiles on each day. Both POAM profiles on Dec 3 show
excellent overall agreement with the MkIV profile, deviating only at the very top of the
profile, above 30 km. The March 15 results are also very satisfying. The POAM profile
measured at 39 E, which we have seen is sampling coincident air masses with the
balloon, reproduces the detailed vertical structure in the MkIV profile almost exactly,

29
whereas the other POAM profile shows a very different vertical structure, consistent with
fact that it is sampling very different air.

In Figure 14 we have plotted the ozone differences for each flight, calculated from
equation (1). The two curves in each panel correspond to the two coincident POAM
profiles on each date, with the red and blue curves having the same meaning as in Figure
13. On December 3 the differences are within 10 % with no obvious bias between 15 and
30 km, which is quite good. Above 30 km and below 15 km there is a systematic
tendency for POAM to be biased high relative to MkIV. The disagreement is 10 – 20 %
above 30 km and larger below 15 km, where it averages 20 % but reaches values as high
as 30 % or more. The POAM satellite validation [Rusch et al., 2001] shows no evidence
of a high bias in the version 3.0 ozone between 30 and 34 km. This is also not a known
systematic bias of the MkIV instrument. However, it is useful to keep in mind that this is
the upper limit of the MkIV retrieval range and the ozone retrievals at or near the balloon
float altitude become very sensitive to assumptions about the shape of the O3 profile
above 33 km. These uncertainties are reflected in the larger MkIV error bars, which
increase from < 0.1 ppm at 30 km altitude to ± 0.4 ppm at 34 km. Therefore the
MkIV/POAM difference at these altitudes is probably not statistically significant.

For the March 15 flight the agreement between MkIV and the closest POAM
measurement at 39 E is within 5 – 10 % at almost all altitudes. The exception is the 15 –
18 km region, where there is some mismatch in the small-scale structure seen in the two
profiles, and the very lowest point at 12 km. Nevertheless, the overall agreement
between these two profiles is very good, particularly when one keeps in mind the fact that
the measurements were made almost 12 hours apart. As expected, the second POAM

30
profile on this day does not agree with the MkIV nearly as well as the first above 15 km,
with a very different vertical structure and differences in the ± 25 % range.

6. DOAS balloon comparisons

A single POAM measurement on February 18 2000 coincided with a balloon
measurement made by the Laboratoire de Physique Molèculaire et Applications and
Differential Optical Absorption Spectroscopy (LPMA/DOAS) instrument. The
LPMA/DOAS balloon gondola was launched from Kiruna into the middle stratosphere
and measured a suite of atmospheric trace gases including ozone. The LPMA/DOAS
gondola carried two optical spectrometers (a DOAS and a FT-IR spectrometer) that
analyze the direct UV-visible and near-IR part of the solar spectrum for line of sight
(LOS) absorption - taken from the balloon gondola to the Sun - for a suite of atmospheric
trace gases (for details of the measurement technique see, e.g., Camy-Peyret et al. [1993]
and Ferlemann et al. [2000]).

The LPMA/DOAS gondola employs two kinds of observations, (1) measurements
made through the atmosphere during balloon ascent for solar zenith angles typically
smaller than 85o and (2) solar occultation observed from balloon float altitude (~ 30 km).
LOS ozone slant column density (LOS-SCD) is inferred by applying the DOAS
technique to direct Sun spectra measured in the ozone Chappuis bands (495.2 - 618.2
nm). Ozone density profiles are then inferred from the measured LOS-SCD values using
either the onion peeling, or the minimal estimate technique. For more details on the
retrieval analysis see Ferlemann et al. [2000].

31
Taking into account the different spectrum integration times and spectral
averaging, the vertical resolution of the measurements is as high as 100 m in ascent mode
and 1 km in solar occultation mode. Absolute accuracy of the ozone retrievals is
estimated to be 2.5 % in ascent mode and 2 % in solar occultation mode. Comparison of
the DOAS ozone with in-situ measurements made by an electrochemical cell (ECC) on
the same gondola agree within the range given by the uncertainty of the individual
measurement (ECC accuracy ± 4 %, DOAS accuracy ± 2 %).

Only one POAM measurement was close enough to the balloon position on this
day to give a reasonable coincidence. Like POAM, the balloon measured a sunset
occultation on this flight and therefore the temporal coincidence is quite good between
POAM and the DOAS occultation measurement (the time of the DOAS ascent
measurement, on the other hand, precedes the POAM measurement by 2 – 3 hours). To
quantify the measurement positions relative to the vortex edge, Figure 15 shows the
equivalent latitude profiles calculated for the DOAS and POAM locations from the
UKMO PV fields. The solid and dashed-dotted blue curves represent the DOAS
occultation and ascent profiles, respectively, and the solid red curve is the POAM profile.
The dashed and dotted black profiles represent the equivalent latitude of the middle and
inner vortex edges, respectively.

These results show that both DOAS measurements occur inside the inner vortex
edge at all altitudes except perhaps at the very bottom of the profile. Above 23 km the
ascent and occultation profiles coincide but below that the ascent measurement samples
higher equivalent latitude than the occultation. The POAM measurement is also
coincident in equivalent latitude with DOAS, and therefore inside the inner vortex edge,

32
above 22 km but samples lower equivalent latitudes than either of the DOAS profiles at
lower altitudes. Due to the proximity to the vortex edge one might expect a fairly
complicated situation, with significant horizontal gradients in the ozone. In particular it
is interesting to note that the POAM and DOAS measurements happen to straddle the
Scandinavian mountains. The LPMA/DOAS probed air masses in the very lee of the
mountain range (with the ascent measurements more eastward than the occultation
measurements) whereas POAM probed air masses mostly westward (upwind) from the
mountains. Hence a pronounced dynamical disturbance of the ozone field is probably to
be anticipated.

In Figure 16 the POAM ozone profile is plotted in the black curve along with the
DOAS ascent and occultation profiles in the blue and red curves, respectively. There is
good qualitative agreement between the three profiles. In particular the distinct notch
structure in the peak of the ozone profile is well captured by both POAM and DOAS,
although the altitude of this structure is more consistent between the POAM and
DOAS/occultation profiles than the DOAS/ascent. Another feature captured well by all
three measurements is the small positive inflection in the profile between 19 and 25 km,
which in this case is more consistent in the POAM and DOAS/ascent data. It is also
impressive how well the DOAS profiles track the POAM profile all the way down to its
lowest point at 9 km.

The ozone difference profiles are plotted in Figure 17. The difference is
calculated from equation (1) after first interpolating the DOAS data onto the standard
POAM altitude grid (the DOAS ascent data have already been smoothed to a 1-km
resolution consistent with POAM). The POAM and DOAS occultation profiles

33
essentially agree to within 10 % at all altitudes between 10 and 30 km. However there
appears to be a fairly systematic bias in the differences, with POAM generally lower than
DOAS by 3 – 5 % from 11 to 18 km, increasing to 5 – 8 % from 18 to 29 km. The
POAM and DOAS ascent differences, on the other hand, appear to be zero-mean with no
significant bias. However, the scatter in the difference profile does show strong altitude
dependence, with absolute differences less than 5 % between 20 and 30 km but
increasing to 20 % below 15 km. Altogether these comparisons are remarkably good
given the expected natural atmospheric variability in the ozone near the vortex edge.

7. OMS/JPL Ozone comparisons

The JPL balloon-borne in situ ozone instrument is virtually identical to the ozone
instrument on the ER-2 aircraft. Ozone is measured in two chambers, one containing
unperturbed air, the other containing air scrubbed of ozone. The ratio of the absorption
of 254-nm radiation (generated by a single mercury lamp) in the two chambers is
measured simultaneously, canceling out lamp intensity fluctuations. This ratio, coupled
with the well-known O3 absorption cross-section and the temperature (controlled and
easily measured), pressure, and path length of the chambers is used to determine the
mixing ratio of O3 in ambient air. The mixing ratio of O3 is measured every second with
an overall uncertainty (accuracy plus precision) of 3 to 5%. Pressure is measured with a
set of calibrated Baratrons with an accuracy of 1%. On the Observations of the Middle
Stratosphere (OMS) balloon flights, in situ measurements are obtained on slow ascent
(300 meters/minute) to 8 mb, a float of about 20 minutes, and a slow descent (150
meters/minute) through most of the stratosphere, followed by a rapid descent on a
parachute.

34

The JPL Ozone instrument flew on two OMS balloon flights launched from
Esrange during SOLVE, on November 19 1999 and March 5 2000. Salawitch et al.
[2001] use the data to quantify ozone loss in the vortex, and provide a discussion of the
location of the balloon flights and comparisons to MkIV remote measurements as well as
ER-2 NOAA Ozone data. Both OMS flights sampled air deep in the vortex.
Unfortunately POAM was not operating on November 19 (due to a safety shutdown for
the Leonid meteor shower) and therefore a direct coincidence with this OMS flight does
not exist. However, POAM did make a measurement very close to Esrange on the
following day (November 20) and this profile has been used as a basis for the
POAM/OMS comparison for the November flight. For the March 5 OMS flight a good
spatial and temporal coincidence with POAM does exist.

Panels (a) and (b) of Figure 18 show the comparison between POAM and the JPL
in situ ozone data for the two flights. The natural vertical coordinate for the JPL Ozone
measurements is pressure, which is measured directly by the instrument as described
above. Since the fundamental POAM vertical grid is geometric altitude, we have used
the UKMO pressure profile (collocated with the time and location of the POAM
measurement) to convert the POAM ozone profile to a pressure grid for the sake of these
comparisons. In these plots the red profile represents the POAM measurement and the
blue dots represent the entire OMS measurement profile (i.e., ascent and descent phases).
In both cases the agreement is excellent, and rivals that seen for in situ - remote
comparisons of instruments flying on the same balloon gondola [Sen et al., 1998].

35
Even though the two measurements are made a day apart, the November
comparison shown in Figure 18 (a) is interesting because the excellent agreement
confirms the very low ozone abundances seen for deep vortex air (~ 3 ppm for altitudes
above 50 mb). As discussed in Salawitch et al. [2001], this relatively low abundance
reflects normal gas phase chemistry under conditions of low solar illumination: the
shorter wavelengths which produce ozone are absent, but slightly longer near-UV
wavelengths which drive photochemical destruction processes are still present. The
"notch" in the balloon profile around 25 mb is a real atmospheric feature, seen in both the
ascent and descent measurements from the balloon. The lower ozone in the notch may
simply reflect different air parcel trajectory history with different solar illumination, and
may be an indicator of variability, on at least small scales, of ozone in the newly forming
vortex. Since POAM should be capable of resolving a feature of this vertical scale, the
fact that it is not evident in the POAM profile suggests that the notch itself is indeed a
localized feature, which does not recur at the time and location of the POAM
measurement (although the ozone abundance measured by POAM is midway between the
lower amount of the notch and the surrounding air).

The good agreement of balloon and POAM low ozone in the mid- to upperstratosphere for observations separated in space and time suggests that these abundances
are representative of a significant portion of the core of the vortex. Since this is air that
will later descend to PSC-induced ozone loss altitudes, the low initial abundances of
ozone have implications for calculations of ozone loss. Also, even though much of the
vortex core is in darkness, displacements to lower latitudes afford an opportunity for a
solar occultation instrument such as POAM to obtain vortex core measurements, which
can be used for calculations of vortex average ozone loss with appropriate data filtering

36
and selection (see Hoppel et al. [2001]).

For the sake of quantitative comparison, we focus on the March 5 flight, where
the OMS and POAM observations were coincident within 1 degree in latitude, 10 degrees
in longitude and 2 hours in time. The results, shown in Figure 18 (b), show excellent
agreement overall in both the magnitude and vertical structure of the ozone profile. The
ability of POAM to discern the very steep falloff in ozone near 30 mb is especially
noteworthy and demonstrates the capability of the instrument to resolve sharp vertical
structures. The ER-2 NOAA Ozone data for this date also lie on top of the JPL balloon
data for the 50-100 mb region (see Figure 5 of Salawitch et al. [2001]). Combined with
the POAM/ER-2 direct comparison on this date (see Figure 10) these results illustrate the
good overall agreement between POAM and in situ measurements.

Figure 18c shows the relative difference profile as a function of altitude. To
calculate this profile, the JPL data were first binned in ln(pressure) then interpolated to
altitude using the POAM/UKMO pressure profile. The difference was then calculated
according to Equation (1). The difference is within 10 %, and often less than 5 %, from
13 to 29 km, but increases to approximately 18 % at the very lowest altitude of 11 km.
Ozone is quite variable in the lower stratosphere, and variations this large are seen in the
JPL Ozone data itself there. In spite of the generally good quantitative agreement it is
obvious that there is a bias in the difference, with POAM higher than JPL Ozone by 3-5
% on average above 12 km. We considered the possibility that this bias is introduced by
errors in the transformation between pressure and altitude grids. However, the UKMO
and JPL Ozone pressure profiles are in good agreement and the use of an alternate
scheme to calculate the differences, in which the JPL measurements are compared

37
directly to POAM in altitude space using the onboard GPS altitude determination from
the OMS gondola, yielded essentially identical results.

It is quite possible that the difference between the OMS and POAM ozone simply
reflects real atmospheric variation for these close, but still separated, measurements. The
differences seen in March between OMS and POAM are not outside the atmospheric
variability, for example, as seen in November between 25-50 mb in the ascent and
descent data from OMS. This suggests that even over the scale of the balloon flight,
there were real atmospheric differences in the vertical profiles of ozone. It is not possible
with only a single comparison to identify the exact cause of the apparent 3-5% bias
between POAM and JPL Ozone, although it is certainly within the instrumental
uncertainties, as well as reasonable atmospheric variability, especially when the very
different nature of the instruments’ spatial sampling is considered.

8. SAOZ balloon comparisons

The Système D'Analyse par Observations Zénithales (SAOZ) sonde is a
lightweight UV-visible diode array spectrometer that measures the absorption of sunlight
by the atmosphere during the ascent (or descent) of the balloon and during sunset (or
sunrise) from float altitude. A simple conical mirror replaces the gondola orientation or
sun tracker systems generally used on large balloon platforms. The balloon version of
the SAOZ instrument is very similar to the one used for ground-based measurements of
total ozone and NO2 [Pommereau et al., 1988]. It is a commercial flat field, 360
grooves/mm, holographic grating spectrometer equipped with a 1024-diode linear array
and an entrance slit of 50 µm. In this arrangement, measurements are made between 290

38
and 640 nm, with an average resolution of 0.8 nm. Ozone is measured in the visible from
450 to 620 nm, where its absorption cross-section is relatively insensitive to temperature.
The SAOZ is equipped with a Global Positioning System (GPS) receiver, which allows
its location to be determined in three dimensions with an uncertainty of ± 150 m.

The spectral analysis and the inversion scheme used for SAOZ are discussed in
Pommereau et al. [1994]. The inversion assumes the scattered light component to be
negligible, an important point because of the use of a 360° conical mirror plus diffuser
instead of a tracker. The random error in the retrieved ozone profiles varies from 0.1% at
30 km to 1 - 2 % at 11 km. Including the maximum expected cross section errors (1%)
and errors in the reference spectrum used in the spectral analysis (0.5%), the total
accuracy of the SAOZ ozone (random + systematic) ranges from approximately 1.5 % at
30 km to 3.4 % at 11 km. The vertical resolution of the measurements is 1.4 km and the
data are sampled at 1-km intervals.

As part of the THESEO 2000 campaign SAOZ instruments made eight flights
from Kiruna in Sweden (68 N, 21 E) or Andoya in Norway (69 N, 16 E) during the 99/00
winter. All of the SAOZ balloon launches were made at local sunset and thus coincide
closely in time with the POAM measurements (generally within one hour). In selecting
SAOZ data to compare with POAM we allow for the fact that the SAOZ measurement
events can cover a significant horizontal distance due to the balloon drift. We therefore
search for POAM coincidences among the SAOZ ascent and occultation measurements
independently on each day. For the comparisons shown here we have adopted a
separation criteria of ± 2 degrees in latitude and ± 4 degrees in longitude. Using this
criteria there are coincidences with POAM on five days: the Andoya flight on Nov 17

39
and the Kiruna flights on January 28, February 27, March 7 and March 25. On February
27 and March 25 only one of the two SAOZ measurements (the occultation and ascent
profiles, respectively) satisfied the coincidence criteria. On all other days both SAOZ
profiles are retained. Despite the stated coincidence criteria, almost all of these SAOZ
measurements occur within 1 degree in latitude and 3 degrees longitude of the POAM
tangent point.

Both the SAOZ and POAM measurements occurred well within the polar vortex
on all these days except for February 27, where the measurements coincided with the
inner vortex edge, and March 25, where both measurements were well outside the vortex.
Figure 19 shows the comparison of the POAM and SAOZ ozone profiles on each of these
days. The black curve represents the POAM profile, while the blue and red curves are
the SAOZ ascent and occultation measurements, respectively. For the most part, the
agreement shown in Figure 19 is very good. Often the SAOZ ascent and occultation
profiles show different vertical structure in the profile, indicating a significant amount of
atmospheric variability even over the temporal and spatial scales spanned by the balloon
measurement. The only notable exception to the good agreement between the two
instruments is the February 27 coincidence, where POAM appears to be systematically
low with respect to the SAOZ occultation profile. While these two measurements are
made in close spatial proximity, as noted previously this is the only occasion where both
POAM and SAOZ are sampling near the vortex edge instead of well inside or outside the
vortex. One might therefore expect a greater degree of natural spatial variability in the
ozone field in this situation

40
Figure 20 shows the mean POAM/SAOZ ozone difference profile, calculated by
averaging over all coincident ascent and occultation profiles shown in Figure 19.
Between 14 and 27 km the absolute difference is within 5 - 7 % everywhere, however
POAM appears to be biased low compared to SAOZ by 2 – 3 % on average. Below 14
km POAM is high relative to SAOZ, with differences increases to a maximum of ~ 20 %
at 10 km.

9. Summary and conclusions

A summary of the POAM ozone comparisons made in this paper is shown in
Figure 21. Here we have only used comparisons that were obtained either completely in
or completely out of the vortex. Measurements made near the edge of the vortex were
discarded because of the possibility of large horizontal gradients in the ozone, which can
significantly complicate interpretation of the results. Thus, we have eliminated the DC-8
comparisons obtained on March 9, and have also eliminated the low altitude (< 15 km)
portions of the January 16 and March 3 DC-8 coincidences. For the ER-2 comparisons,
we have used the direct coincidence comparisons discussed in Section 4.3 because they
cover the largest altitude range. For the MkIV comparisons we have averaged the
differences obtained from the two coincident profiles on December 3, as well as the
single coincidence at 39 E on March 15, which we feel is a valid coincidence despite the
vortex edge conditions (see Figure 12 and accompanying discussion). For the
POAM/DOAS summary we have averaged the differences from the DOAS occultation
and ascent profiles. Again, we feel these comparisons are valid in spite of their proximity
to the vortex edge, based on the arguments made previously. The OMS result comes
from the single coincidence on March 5 2000 and is identical to the result shown in

41
Figure 18c. Finally, the SAOZ profile is obtained by averaging all the coincidences
shown in Figure 19 except for the February 27 event, which occurred at the vortex edge
as discussed in Section 8.

The comparisons shown in Figure 21 naturally divide into 3 altitude regions:
below 14 km, 14-30 km, and above 30 km.

In the primary altitude range between 14 and

30 km, where the majority of the coincident measurements are made and the statistics are
therefore the best, POAM agrees with all the SOLVE data sets examined in this paper to
within 7-10 % with no apparent bias (this conclusion ignores the divergence with UV
DIAL above 20 km, which is explained in section 3.3, and the peak in the
POAM/AROTEL difference at 28-29 km, which is dominated by a single bad
comparison on December 2). The observed differences are within the combined errors of
POAM and the correlative measurements in this altitude range and collectively
demonstrate an impressive degree of consistency between the various datasets, despite
the very different measurement techniques and spatial sampling inherent in the data.
These differences are also consistent with the POAM satellite and ECC sonde validation
in this altitude range [Rusch et al., 2001].

Below 14 km there is a great deal of scatter in the SOLVE comparisons. The
most extensive comparisons come from the ER-2 measurements, which indicate a high
POAM bias (~ 20 %) at the lower altitudes, but only below 10 km. The SAOZ and MkIV
comparisons also tend to show POAM biased high by 10–20 % below 14 km. However
the UV DIAL, DOAS, and JPL/OMS comparisons, while exhibiting larger differences in
this altitude range ( ± 15 %), show no consistent bias. Validation of the POAM version
3.0 ozone with satellites and ECC sondes does indicate a high bias in the Northern

42
Hemisphere below 14 km but the magnitude is somewhat uncertain, ranging from 5 to
25% at 10 km [Rusch et al., 2001]. In summary, it is difficult to evaluate the consistency
of the SOLVE comparisons in this altitude range with previous comparisons, and the
actual POAM accuracy remains somewhat uncertain.

Above 30 km there are only the AROTEL and MkIV comparisons, both of which
seem to suggest a high POAM bias in this altitude range. However, there is a large
amount of scatter in these results, which are based on only 4 measurements (the
AROTEL coincidences on December 14 and March 9 and the two MkIV coincidences on
December 3). It is worth noting that comparisons with HALOE and SAGE II do not rule
out the possibility of a high POAM bias between 30 and 35 km, but the magnitude of the
discrepancy is no more than 5 % in the satellite comparisons [Rusch et al., 2001].
Therefore, we feel that the SOLVE comparison results are inconclusive above 30 km.

In conclusion, the SOLVE campaign has provided abundant opportunities for
validating POAM ozone. This is especially true in the 14 - 30 km region. In this altitude
range the comparisons with all seven correlative data sets considered in this study are
remarkably consistent, and indicate agreement to well within the expected uncertainties
in POAM ozone obtained both from formal error analysis [Lumpe et al., 2001] and other
validation studies [Rusch et al., 2001]. However, the situation is not as clear-cut below
14 km or above 30 km where few measurements and large scatter complicate the
comparisons. The altitude region above 30 km is not an important issue because it is
very clear that, in this altitude range, ozone can be measured very well from satellites,
and validation opportunities (apart from those obtained in dedicated campaigns) abound.
The altitude region below 14 km is much more important. In this region, for a variety of

43
reasons, ozone is much more difficult to measure by satellite-based instruments, and the
accuracy of these measurements is not very clear. However, long-term data sets provided
by satellites in this altitude region have significant scientific value, so their reliability
must be carefully assessed. Obtaining a large number of high-precision measurements in
this altitude region should be a very high priority in the planning of future satellite
validation campaigns.

Acknowledgments

This work was supported by the NASA Scientific Data Purchase program through
contract # NAS13-99031. The version 3.0 POAM data are archived at the NASA
Langley Research Center EOSDIS Distributed Active Archive Center. We wish to thank
Ross Salawitch for his work in making the entire ER-2 data set available in a very
convenient form in terms of his “merge files”.

The balloon experiments have been operated by the teams of the NASA National
Scientific Balloon Facilites (NSBF), the Centre National d'Etude Spatiales (CNES), the
Swedish Space Corporation (SSC) and the Norwegian Space Center (NSC) which are
gratefully acknowledged. European SAOZ and LPMA/DOAS experiments were
supported by the EC Environmental Programmes THESEO-Eurosolve EVK2-199900311 and Alomar LSF.

44

References

Browell, E. V., Differential absorption lidar sensing of ozone, Proc. IEEE, 77, 419-432, 1989.

Browell, E. V., C. F. Butler, S. Ismail, M. A. Fenn, S. A. Kooi, A. F. Carter, A. F. Tuck,
O. B. Toon, M. H. Proffitt, M. Loewenstein, M. R. Schoeberl, I. Isaksen, and G.
Braathen, Airborne lidar observations in the wintertime Arctic stratosphere: Ozone,
Geophys. Res. Lett., 17, 325-328, 1990.
Browell, E. V., C. F. Butler, M. A. Fenn, W. B. Grant, S. Ismail, M. Schoeberl, O. B.
Toon, M. Loewenstein, and J. R. Podolske, Ozone and aerosol changes observed
during the 1992 Airborne Arctic Stratospheric Expedition, Science, 261, 1155-1158,
1993.
Browell, E. V., S. Ismail, and W. B. Grant, Differential absorption lidar (DIAL)
measurements from air and space, Appl. Phys. B 67, 399-410, 1998.
Camy-Peyret, C., Flaud, A. Perrin, C.P. Rinsland, A. Goldman, and F. Murcray,
Stratospheric N2 O5 , CH4 , and N2 O profiles from IR solar occultation spectra, J.
Atmos. Chem., 16, 31 - 40, 1993.
Danilin, M. Y., et al., Comparison of ER-2 aircraft and POAM III, MLS, and SAGE II
satellite measurements during SOLVE using traditional correlative analysis and
trajectory hunting technique, submitted to J. Geophys. Res, 2001.
Farmer, C. B., High-resolution infrared spectroscopy of the Sun and the Earth’s
atmosphere from space, Microchim. Acta [Wien], III, 189-214, 1987.
Ferlemann, F., N. Bauer, R. Fitzenberger, H. Harder, H. Osterkamp, D. Perner, U. Platt,
M. Schneider, P. Vradelis, and K. Pfeilsticker, Differential Optical Absorption

45
Spectroscopy Instrument for stratospheric balloon-borne trace gas studies, Applied
Optics, 39, 2377 - 2386, 2000.
Grant, W. B., M. A. Fenn, E. V. Browell, T. J. McGee, U. N. Singh, M. R. Gross, I. S.
McDermid, L. Froidevaux, and P.-H. Wang, Correlative stratospheric ozone
measurements with the airborne UV DIAL system during TOTE/VOTE, Geophys.
Res. Lett., 25, 623-626, 1998.
Hoppel, K. et al., POAM III observations of Arctic ozone loss for the 1999/2000 winter,
submitted to J. Geophys. Res., this issue, 2001.
Lucke, R.L., et al., The Polar Ozone and Aerosol Measurement (POAM) instrument and
early validation results, J. Geophys. Res, 104, 18,785-18,799, 1999.
Lumpe et al., POAM III retrieval algorithm and error analysis, submitted to J. Geophys.
Res., 2001.
McGee, T. J., M. Gross, U.N. Singh, J. J. Butler, and P. Kimvilakani, "An Improved
Stratospheric Ozone Lidar", Opt. Engin., 34, 1421-1430, 1995.
McGee, T. J., J. F. Burris, L. Twigg, D. Silbert, W. Heaps, W. Hoegy, G. Sumnicht, C.
Hostetler, G. Hansen, P. Lucker, M Osborne, M Lawrence, R. Neuber and T.
Schmidt, The AROTEL Instrument - an airborne lidar for stratospheric ozone,
aerosols and temperature, submitted to Applied Optics, 2001.

Nash, E. R., P. A. Newman, J. E. Rosenfield, and M. R. Schoeberl, An objective
determination of the polar vortex using Ertel’s potential vorticity, J. Geophys. Res.,
101, 9471-9478, 1996.

Pommereau, J. P. and J. Piquard, Ozone, nitrogen dioxide and aerosol vertical
distributions by UV-visible solar occultation from balloons, Geophys. Res. Lett, 21,
1227-1230, 1994.

46
Pommereau, J.P. and F. Goutail, O3 and NO2 ground-based measurements by visible
spectrometry during Arctic Winter and Spring 1988, Geophys. Res. Lett, 891, 1988.

Proffitt, M. H., et al., In situ ozone measurements within the1987 Antarctic ozone hole
from a high-altitude Er-2 aircraft, J. Geophys. Res., 94, 16,547-16,555, 1989.

Randall, C.E., R.M. Bevilacqua, J.D. Lumpe, K.W. Hoppel, D.W. Rusch, and E.P.
Shettle, Comparisons of Polar Ozone and Aerosol Measurement (POAM) II and
Stratospheric Aerosol and Gas Experiment (SAGE) II aerosol measurements from
1994 to 1996, J. Geophys. Res. 105, 3929-3942, 2000.

Randall, C.E., R.M. Bevilacqua, J.D. Lumpe and K.W. Hoppel, Validation of POAM III
aerosols: Comparison to SAGE II and HALOE, Validation of POAM Aerosols:
Comparison to HALOE and SAGE II, submitted to J. Geophys. Res., 2001a.

Randall C. E. et al., Construction of 3D Ozone Fields Using POAM During SOLVE,
submitted to J. Geophys. Res., this issue, 2001b.

Rusch et al, "Validation of POAM II Ozone Measurements with Coincident MLS,
HALOE, and SAGE II Observations", J. Geophys. Res., 102, 23,615-627, 1997.

Rusch, D.W., et al., Validation of POAM III O3: Comparison to ozonesonde and satellite
data, submitted to J. Geophys. Res., 2001.

Salawitch, R. J., J. J. Margitan, B. Sen, G. C. Toon, G. B. Osterman, M. Rex, J. W.
Elkins, E. A. Ray, F. L. Moore, D. F. Hurst, P. A. Romashkin, R. M. Bevilacqua, K.
Hoppel, E. C. Richard, and T. P. Bui, "Chemical Loss of Ozone During The Arctic

47
Winter of 1999-2000: An Analysis Based on Balloon-borne Observations", submitted
to J. Geophys. Res., this issue, 2001.

Sen, B., G. C. Toon, G. B. Osterman, J.-F. Blavier, J. J. Margitan, R. J. Salawitch, and G.
K. Yue, Measurements of reactive nitrogen in the stratosphere, J. Geopys. Res., 103,
3571-3585, 1998.

Toon, G. C., The JPL MkIV Interferometer, Opt. Photonic News, 2, 19-21, 1991.

48

Figure Captions

Fig. 1 Comparison of the POAM (solid line) and predicted SAGE III (dashed line)
Northern Hemisphere measurement latitudes during the year. For POAM this is
invariant from year to year. The SAGE III ephemeris will likely change slightly
depending on exact launch date and time.

Fig. 2 Locations of the POAM and DC-8 coincidences for six DC-8 flights. The blue line
on each map represents the DC-8 flight track. The red circle is the location of the
POAM measurement for which the coincidence comparisons are made. It is
surrounded by a red rectangle that represents the area defined by the coincidence
criteria ( ± 1 deg latitude, ± 2 deg longitude). Black contours (solid, dashed, solid)
represent the vortex inner, middle and outer edge, respectively, on the 450 K
potential temperature surface.

Fig.3 POAM (black), UV DIAL (red) and AROTEL (blue) ozone profiles measured
during the six coincidence periods shown in Figure 2. The POAM curves
correspond to the single profile measured at the location represented by the red
circle in Figure 2. All UV DIAL and AROTEL data meeting the coincidence
criteria for each flight have been averaged to produce the single profile shown here.

Fig. 4 Ozone difference, as calculated by equation (1), for POAM/UV DIAL (top panel)
and POAM/AROTEL (middle panel). Blue and red curves represent the difference

49
profiles for each of the 6 flights, whereas the black curve is the mean difference of
all flights. Error bars correspond to the standard error of the mean difference. The
bottom panel simply compares the POAM/UV DIAL and POAM/AROTEL mean
error profiles from the top two panels.

Fig. 5 Examples of POAM and ER-2 data used in the vortex-averaged ozone
comparisons for 4 ER-2 flights. The blue curves correspond to the ER-2 flight track
on each day. The black and green contours represent the location of the inner and
middle vortex edge, respectively, on the 450 K potential temperature surface. Red
circles show the location of all POAM measurements made in the three-day period
centered on the ER-2 flight date. Only those POAM and ER-2 measurements made
inside the black contour are used in the vortex average comparisons.

Fig. 6 Vortex-average ozone mixing ratio profiles measured by POAM (red circles) and
ER-2 (black squares) for the 11 ER-2 flights indicated at the top of each panel. The
solid black and red lines represent the standard error of the means for POAM and
ER-2, respectively.

Fig.7 POAM (red circles) and ER-2 (black triangles) ozone measurements at 460 K on
Jan 23 (top panel) and March 11 (bottom panel). The POAM data include all
measurements made within ± 1 day of flight date. Only in-vortex ER-2 data are
plotted. The vertical lines denote the middle and inner vortex edges.

50
Fig. 8 The left panel shows the POAM and ER-2 in-vortex ozone mixing ratio profile
averaged over all 11 ER-2 flights shown in Figure 6 (symbols are the same as in
Figure 6). The right panel shows the mean ozone difference profile for the 11
flights. Error bars represent the standard error of the mean difference.

Fig. 9 Results of the trajectory matching analysis for 11 ER-2 flights. Black dots
represent the ozone difference for all trajectory parcels that satisfy the coincidence
criteria outlined in section 4.2. There are 249 such matches. The black curve is the
mean difference at each potential temperature level and error bars represent the
standard error of the mean difference.

Fig.10 POAM (blue) and ER-2 (red) ozone mixing ratios satisfying the coincidence
criteria on each of 10 ER-2 flights. The coincidence criteria used are ± 3 deg
latitude, ± 10 deg longitude and ± 3 hours. The POAM profiles represent the single
coincident measurement from each day. The ER-2 profiles are constructed by
binning all coincident ER-2 data for each flight into 1-km altitude bins.

Fig. 11 Summary of POAM /ER-2 ozone differences obtained from three comparison
techniques. The black curve represents the mean difference obtained from the
POAM/ER-2 direct coincidences on the 10 flights shown in Figure 10. The number
of coincident ER-2 data points at each altitude level is shown at the right edge. For
comparison the blue and red curves are the mean differences from the vortexaveraging and trajectory matching analyses, respectively (see Figures 8 and 9).
These have been interpolated to an altitude grid using a mean vortex-averaged

51
potential temperature profile. In all cases the error bars represent the standard error
of the mean difference.

Fig. 12 Equivalent latitude profiles for the coincident POAM (red) and MkIV (blue)
ozone measurements made on Dec 3 1999 and March 15 2000. For comparison the
equivalent latitude of the inner and middle vortex edges are plotted in the black
curves.

Fig. 13 Coincident ozone mixing ratio profiles measured by MkIV (black curves) and
POAM (blue and red curves) on Dec 3 1999 and March 15 2000. The two POAM
profiles correspond to the two closest measurements to the balloon position on each
day.

Fig.14 Ozone difference profiles (from equation (1)) for the POAM /MkIV coincidence
measurements shown in Figure 13. For each day the red and blue curves represent
the difference between MkIV and the POAM profile of the same color in Figure 12.

Fig. 15 Equivalent latitude profiles for the coincident POAM (red) and DOAS (blue)
ozone measurements made on February 18 2000. For comparison the equivalent
latitude of the inner and middle vortex edges are plotted in the black curves.

Fig. 16 Coincident ozone density profiles measured by POAM (black curve) and DOAS
ascent mode (blue curve) and occultation mode (red curve) on February 18 2000.

52

Fig. 17 Ozone difference profiles (from equation (1)) for the POAM /DOAS coincidence
measurements shown in Figure 16. The red and blue curves represent the
difference between the POAM measurement and the DOAS occultation and
ascent measurements, respectively.

Fig. 18 Panel (a) shows the ozone mixing ratio profile measured by the JPL Ozone
instrument on November 19 2000 in the blue symbols. The red curve represents
the ozone measured by POAM on November 20 at approximately the same
location (see figure legend for exact measurement locations). Panel (b) shows
coincident ozone mixing ratio measurements made by POAM (red curve) and JPL
Ozone (blue symbols) on March 5 2000. Panel (c) represents the ozone difference
profile calculated for the March 5 2000 coincidence.

Fig. 19 Ozone density profiles measured by POAM (black curve) and SAOZ ascent
mode (blue curve) and occultation mode (red curve) for coincident events on five
different days. The coincidence criteria used are discussed in Section 8. The
locations of the measurements are listed in the legend, with the text color chosen
to correspond to the ozone profiles.

Fig. 20 Mean ozone difference profile (from equation (1)) for the POAM /SAOZ
coincidence measurements shown in Figure 20. Error bars correspond to the
standard error of the mean difference.

53
Fig. 21 Summary comparison of ozone difference between POAM and AROTEL (solid
blue), UV DIAL (solid red), ER-2 (solid green), MkIV (black), DOAS (dashed
blue), OMS (dashed red) and SAOZ (dashed green). All profiles represent the
mean difference calculated from equation (1). Only coincident measurements
obtained either completely in or completely out of the vortex are used to calculate
these final difference profiles, as discussed in Section 9.

54

Figure 1

55

Figure 2

56

Figure 3

57

Figure 4

58

Figure 5

59

Figure 6

60

Figure 7

61

Figure 8

62

Figure 9

63

Figure 10

64

Figure 11

65

Figure 12

66

Figure 13

67

Figure 14

68

Figure 15

69

Figure 16

70

Figure 17

71

Figure 18

72

Figure 19

73

Figure 20

74

Figure 21