Comparisons of remote sensing stratospheric mixing ratio profiles for O3, H2O, N2O, HNO3, and NO2 are shown between data (version 5.2) obtained by the Improved Limb Atmospheric Spectrometer (ILAS) on board the Japanese Advanced Earth Observing Satellite (ADEOS) and the Smithsonian Astrophysical Observatory far-infrared spectrometer (FIRS)-2 balloon-borne spectrometer from a flight on 30 April 1997 originating in Fairbanks, Alaska. Submillimeter wave remote sensing and UV in situ observations of mixing ratios of O3 from Jet Propulsion Laboratory (JPL) instruments obtained on board the same balloon gondola are also used for comparison with ILAS mixing ratios of O3. The remote sensing balloon observations occurred roughly 5° latitude, and the in situ 2° latitude, north of the nearest ILAS observations and were taken between 0700 and 1300 local solar time (LST), while the ILAS data were taken at the previous local sunset (near 2100 LST). Back trajectories from the locations and times of balloon data sets to the prior local sunset were within 2° latitude and 10° longitude of the nearest ILAS occultation; the validation therefore effectively occurred in very nearly the same air mass as was observed with ILAS, and well within a 1-day time interval. The mixing ratio profiles of all of the compared molecules (O3, HNO3, N2O, H2O, and NO2) agree to within the combined uncertainties with only minor systematic offsets.
 The Japanese Advanced Earth Observing Satellite (ADEOS) was a multi-instrument platform designed to examine terrestrial parameters. The ADEOS platform was on a Sun-synchronous polar orbit and operated continuously and routinely from 30 October 1996 to 29 June 1997. The mission of two of the instruments was to monitor the Earth's atmosphere. One of them, the Improved Limb Atmospheric Spectrometer (ILAS) instrument, probed the stratosphere by making high latitude observations in both the southern (64–88°S) and northern (57–73°N) hemisphere.
 Two balloon instrument based validation campaigns were undertaken to overlap with ILAS observations, one from Kiruna, Sweden and one from Fairbanks, Alaska. The latter campaign occurred in April and May of 1997 and involved two large balloon payloads. One payload carried the Smithsonian Astrophysical Observatory (SAO) far-infrared spectrometer (FIRS)-2, the Jet Propulsion Laboratory (JPL) Submillimeter Limb Sounder (SLS), and an in situ UV photometer (also from JPL) for concentrations of O3. The other payload carried the JPL MkIV interferometer and University of Denver CAESR grating spectrometer. All of these instruments previously had been involved in the validation of the Upper Atmosphere Research Satellite (UARS) data [Gille et al., 1996].
 Here, we will discuss the comparisons between ILAS version 5.2 data and observations from the balloon payload containing the FIRS-2, SLS, and in situ instruments for the molecules O3, HNO3, N2O, NO2, and H2O. We discuss the degree of coincidence of these sets of observations, as well as possible reasons for some of the observed discrepancies. Comparisons with ILAS version 5.2 data using observations from the MkIV balloon flight are contained in a companion paper [Toon et al., 2002].
 Papers describing the comparisons for all correlative observations of individual parameters with ILAS (including all balloon campaigns, ozone sondes, aircraft data, and ground based data) have been prepared that include all the observations included here [Kanzawa et al., 2002; Yokota et al., 2002; Sugita et al., 2002; Irie et al., 2002]. Here, we concentrate on the robustness of the overlap of the FIRS-2, SLS, and in situ ozone balloon observations with ILAS, the systematic uncertainties of the data sets, and discuss possible causes for the discrepancies that are consistent with all of the molecules being compared. Such a comparison is not possible with a study that focuses on the comparison of multiple data sets one molecule since the different data sets will have different systematic uncertainties and different degrees of spatial and time overlap with the ILAS data.
 The ILAS instrument, developed at the Environmental Agency of Japan, is a grating solar occultation spectrometer that measures atmospheric absorption in limb viewing geometry [Sasano et al., 1999a]. It contains 2 spectrometers, one that measures infrared wavelengths between 6.21 and 11.77 μm with a 44 element pyroelectric detector array and one that measures visible wavelengths between 753 to 784 nm with a 1024 element detector array. The latter wavelength range contains the O2 A bands for retrieval of temperature and possible refinement of observation geometry and is used to quantify extinction from aerosols. The IR wavelength region contains significant atmospheric absorption from O3, H2O, N2O, NO2, HNO3, and CH4. Broad and minor, but detectable, absorptions also occur from CFC-11, CFC-12, and aerosols. Mixing ratio profiles for these molecules are obtained from near the tropopause to as high as 60 km, depending on the molecule. ILAS observations in the northern hemisphere are taken in sunrise mode during the spring which corresponds to local solar sunset. A more complete description of the instrument is given by Sasano et al. [1999a] and Nakajima et al. [2002b], and a full description of the retrieval algorithm for version 5.2 data can be found in this issue [Yokota et al., 2002; Nakajima et al., 2002a]. One should refer to these articles for a complete description of the algorithm used here. The uncertainties listed here include random errors from spectral noise, correlation errors between the retrievals of the different parameters, and pointing uncertainties.
 The FIRS-2 is an FTIR spectrometer that measures atmospheric thermal emission at long wavelengths in the infrared, between 6 and 120 μm, with a spectral resolution of 0.004 cm−1 [Johnson et al., 1995]. The wavelength range is larger than for the previous balloon flights of FIRS-2 due to a new beam splitter design used for this flight [Dobrowolski and Traub, 1996]. When deployed from balloon platforms, FIRS-2 obtains spectra from balloon float altitudes (roughly 37 km) in a limb viewing geometry down to tangent altitudes near the tropopause. From these spectra, mixing ratio profiles are obtained for 28 molecules, including O3, H2O (4 isotopomers), CO2, N2O, HCl, HOCl, ClNO3, NO2, HNO3, HNO4, OH, HO2, H2O2, CFC-11, CFC-12, and CH4. Past FIRS-2 observations have been used to validate many of the mixing ratio profiles retrieved by instruments on board the UARS [Kumer et al., 1996; Roach et al., 1996; Lahoz et al., 1996; Russell et al., 1996a, 1996b; Harries et al., 1996]. All the uncertainties shown in the plots below contain random retrieval errors from spectral noise and the systematic errors from uncertainties in atmospheric temperature and limb pointing angle. They do not include spectroscopic uncertainties which are discussed separately for each molecule since they affect the comparisons differently.
 The SLS instrument is a submillimeter heterodyne remote sensing spectrometer that retrieves profiles of ClO, O3, and HCl [Stachnik et al., 1992]. Of these, ILAS only measures O3. The JPL in situ O3 instrument is a high-precision UV photometer that measures O3 locally during the duration of a balloon flight. It has a precision of 3 ppt and an overall accuracy of 3% [Hilsenrath et al., 1986]. Past SLS and in situ observations of [O3] have been used to validate the MLS profiles on board the UARS [Froidevaux et al., 1996].
 The balloon flight carrying the FIRS-2, SLS, and UV in situ photometer originated from Fairbanks, Alaska on 30 April 1997 as part of the Alaska Balloon Campaign which was undertaken for the validation of ADEOS instruments. The location of the nearest ILAS occultation to location and time of the balloon observations occurred at 63.6°N, 210.3°E during previous local sunset. The UV photometer obtained a mixing ratio profile for O3 on ascent of the balloon. The average location of the O3 in situ profile on ascent was 65.1°N, 212.2°E. The SLS and FIRS-2 obtained profiles while at float altitude in limb viewing geometry, at the same azimuths, first observing to the north of the balloon for 2 hours, then to the northwest of the balloon for almost 3 hours. The flight lasted nearly 5 hours at float altitude. The locations of the average target point for these two observation geometries were 69.7°N, 211.0°E and 68.0°N, 200.8°W, respectively.
Figure 1 shows the locations of balloon observations on 30 April 1997 relative to the nearest ILAS observation on 29 April 1997 (which corresponds to 0604 UT on 30 April 1997). The times and locations of the FIRS-2 and SLS points in Figure 1 are the average for observations made with two different azimuth pointing directions used during the balloon flight. Since the balloon observations were made well after the sunset occultation observed by ILAS, we have calculated back trajectories for parcels from the balloon observation points at 20 and 32 km back to the time of the ILAS observation (local sunset, 12 to 15 hours) using the NMC trajectory analysis via automail from the NASA Goddard Space Flight Center [Schoeberl and Sparling, 1994]. The trajectory end points for the balloon observations at 20 km fall within 2° latitude of the ILAS observations. Two of the trajectory end points fall within 3° longitude of the ILAS occultation. At 32 km, the trajectory endpoints all fall within 3° latitude of the ILAS observations, and are roughly 10° east of the ILAS occultation. The ILAS occultation at the sunset on 30 April 1997 was at 63.5°N and 217.9°W but projecting forward in time from the balloon observations to that sunset produces worse spatial overlap between balloon and ILAS observations, with the balloon locations projecting to about 8° north of the ILAS occultation and 10° west.
 The Alaska Balloon Campaign occurred well outside of the arctic vortex which was still intact at the time of the two balloon flights. The effective sample location of all measurements reported here also occurred well outside the vortex. Potential vorticity gradients near the ILAS and balloon observations were small. As a result of this and the above trajectory analysis, we don't expect significant variability between the small displacements of the observation locations of the ILAS and balloon data, and expect the Alaska balloon campaign comparisons to be geographically and meteorologically robust for validating ILAS data. Correlative measurements taken near the edge or just inside the arctic vortex may have large gradients over small distances, making comparisons more difficult to quantify than in the case presented here.
 ILAS and FIRS-2 use different spectroscopic features for the retrievals of concentrations for many of the above molecules. FIRS-2 operates in thermal emission throughout the far infrared and into the mid infrared region while ILAS operates in solar occultation mostly in the mid infrared. The only molecules where the same spectroscopic features are used for the two instruments are CFC-11, CFC-12, and CH4. For this particular flight, the FIRS-2 spectra in the CH4 spectral region were insufficiently calibrated for CH4 to be used for validation. Thus, different spectroscopic features are used for the two data sets for all of the molecules being compared here and relative errors in spectral line parameters will affect the comparisons.
 Ozone is the common stratospheric constituent measured by ILAS, FIRS-2, and SLS and the in situ photometer (the instruments on the balloon payload being discussed here). Figure 2 shows the comparisons of the nearest concentration profile of O3 from ILAS (v5.2) with all three instruments on board the balloon gondola. The spectroscopic uncertainties for O3 for each of these instruments are less than 5%. As a result, these uncertainties are not added to any of profiles shown in this figure. There is agreement to within the 1 sigma uncertainties between both sets of FIRS-2 profiles, both sets of SLS profiles, and the in situ profile above 20 km. Below 20 km, while there is variability between the balloon data sets by as much as 30%, the variations are consistent with the structure shown in the in situ profile (roughly 0.5 ppm which is 25% to 50% of the mixing ratio at these altitudes), suggesting significant spatial variability in the observed air masses. Here, the FIRS-2 mixing ratios are larger than the SLS values by amounts larger than twice the one sigma uncertainties given for the profiles. The cause of this discrepancy is unknown since they are measuring the same air masses. The ILAS concentration profile of O3 falls within the spread of the balloon data sets over the entire altitude range. And, with the exception of the 35–40 km range, the ILAS data are consistent with all the validation data sets to within the uncertainties, and residuals less than 1 ppm. Between 35 and 40 km, the ILAS data is more consistent with the in situ data than the 2 remote sensing data sets where the residuals are about 0.5 ppb with the FIRS-2 data and about 1 ppb with the SLS data. This may result from a significant latitudinal gradient of O3 in the upper stratosphere where the ozone photochemical lifetime (roughly 1 day) is short compared to transport times (30 days). Such a latitudinal gradient at these altitudes is evident in other data sets, such as HALOE (HALogen Occultation Explorer), at these altitudes and latitudes. There is not a systematic altitude dependence for the differences between the ILAS and the balloon observations, suggesting no significant differences in systematic uncertainties in the retrievals resulting from using different spectroscopic features.
 Two recent studies validating earlier versions (v3.0 and v3.1) of ILAS O3 showed comparison between the ILAS concentrations and the validation data sets that are very similar to the comparisons shown here with version 5.2 [Jucks et al., 1998; Sasano et al., 1999b]. This good comparison remains even though major changes occurred in the data processing algorithms between version 3.1 and v5.2, the most significant being a change in the determination of the tangent height registration. For the earlier algorithms, the height registration was performed by matching the equivalent width in the P-branch of the O2 A band in the visible channel to calculated values. Since that time, it was determined that there are uncertainties in the instrument line shape that can adversely affect interpretation of the visible channel of ILAS. For version 5.2, the tangent height is determined directly from the sun edge sensor, as described by Nakajima et al. [2002a].
Sugita et al.  present a compete comparison of many of the O3 validation data sets with the ILAS over the duration of the ILAS lifetime. The Sugita study includes data from the presented paper as well as data from other balloon flights, ozonesondes, and other ozone retrieving satellite instruments such as HALOE and SAGE-II. Their findings are similar to what is shown here in Figure 2, that the ILAS O3 agree with the large set of validation data to within the uncertainties of the ILAS data and to better than 20% at all altitudes, and there doesn't appear to be any systematic offset of the ILAS data as a function of altitude.
 Most of the other stratospheric constituents retrieved by ILAS are also retrieved by FIRS-2. A summary of the comparisons of ILAS version 5.2 data and FIRS-2 data for HNO3, N2O, H2O, and NO2 is shown in Figure 3. The left side shows the mixing ratio profiles while the right side shows the percent difference between the ILAS and two FIRS-2 profiles. The locations and times of these profiles are the same as those in Figure 2 and described above. Three of the molecules shown in Figure 3 (HNO3, N2O, and H2O) have uncertainties of less than 25% while NO2 has an uncertainty of 50% or more.
 The spectroscopic uncertainties for H2O are less than 10% for both FIRS-2 and ILAS. FIRS-2 concentrations of H2O are retrieved from rotational transitions in the far infrared and the estimated uncertainties for the transitions used is 3%. ILAS concentrations are obtained from the ν2 band near 6 microns in the short wavelength channels of its infrared array. Its spectroscopic uncertainties are less than 10%, and probably closer to 5%. Errors in the spectroscopic parameters for either of these two bands or differences in the handling of the water vapor continuum between the data reduction algorithms used can easily account for up to 10% differences between these retrievals. There is agreement to within the uncertainties between FIRS-2 and ILAS profiles of [H2O] throughout most of the altitude range of comparison. Below 20 km, the ILAS values are lower than the second FIRS-2 profile by more than 1 sigma, which is about 10% at these altitudes. This latter profile samples below the hydropause whereas the ILAS data does not. In general, the ILAS data are lower than the FIRS-2 data by 5–10% below 30 km, which is consistent with the combined uncertainties, which range from 10% to 20%, and is on the order of the spectroscopic uncertainties. The comparison with MkIV data also shows this bias but to a larger extent; the ILAS data are systematically lower than the MkIV data by 0.5 to 1 ppm [Toon et al., 2002]. This comparison of data is part of a larger comparison of ILAS H2O retrievals with validation data sets [Kanzawa et al., 2002]. These comparisons shown here in Figure 3a are similar to those found with other validation data sets, showing agreement to within the uncertainties, but the ILAS data being biased somewhat low.
 The comparisons for profiles of HNO3 show agreement to within the uncertainties between 15 and 40 km, with residuals of less than 1 ppb. These comparisons are better than the comparisons of version 3.1 ILAS concentrations of HNO3 with a number of validation data sets (including FIRS-2), described recently [Koike et al., 2000]. In that study, the ILAS values were consistently lower than the validation data sets, especially between 20 and 30 km. The current version ILAS values are higher than FIRS-2 concentrations above 34 km by nearly the 1 sigma uncertainty of the ILAS precision. The improvements here result mostly from the corrections to the pointing algorithm described by Nakajima et al. [2002a]. These observations agree well despite being retrieved from different spectroscopic features; FIRS-2 uses the ν9 band while ILAS uses the ν5/2ν9 bands. FIRS-2 observations using the ν5/2ν9 bands compared well with those from the ν9 band, with the latter being higher by 7 ± 3%. If the ν5/2ν9 bands were used, the difference between ILAS and FIRS-2 would be somewhat smaller between 20 and 30 km, but larger above 30 km. Below 15 km, the ILAS data are roughly 50% of the FIRS-2 data, which is much larger than the uncertainties.
 The ILAS N2O concentrations agree with the FIRS-2 values from 13 to 28 km to within the uncertainties of the ILAS data (roughly 20% to 30% over that altitude range). Below 15 km, the ILAS data are less than the FIRS-2 data by about 30%. At and above 30 km, the ILAS data are greater than the FIRS-2 data by 30–50%, which is the uncertainty at 30 km, but less than the uncertainty above that altitude. This comparison shows a significant improvement over that with previous versions of ILAS data. ILAS retrievals of N2O have been improved considerably from previous versions because of the advances in the altitude registration and the inclusion of CF4 in the radiative transfer calculations.
 The ILAS observations of NO2 are systematically higher than FIRS-2 observations in the lower stratosphere by roughly 30%. The FIRS-2 observations were taken during late morning to near local noon whereas the ILAS data were taken at local sunset. Concentrations of NO2 should be lower at the times of the FIRS-2 observations than at sunset by roughly 10–20%, the time of the ILAS observations, owing to photolysis of N2O5 that proceeds slowly during the day. Considering this diurnal variation in NO2, the discrepancy between the FIRS-2 and ILAS values would be less than 10% near 25 km, but still rather large below 20 km. The [NO2] from FIRS-2 in the middle and upper stratosphere show some variation but agree to within the large uncertainties. The transitions used for the retrieval of [NO2] from FIRS-2 spectra are very weak, causing the relatively large uncertainties and variations shown here. The atmospheric absorption lines used to retrieve [NO2] from the ILAS spectra are also relatively weak at the resolution of ILAS, and contribute mostly in the short wavelength spectral element. These retrievals will be highly correlated with contributions from H2O and aerosols. Errors in accounting for the absorption of these two components in the last element could explain the large discrepancies seen in the lower stratosphere.
 All the molecules discussed here show generally show agreement to within the uncertainties between the mixing ratio profiles for ILAS and FIRS-2. This suggests that the data analysis algorithm used for the ILAS spectra is appropriate and that the current altitude registration algorithm used is correct. As a result, we find that systematic uncertainties in the version 5.2 of ILAS data sufficiently small to make this data useful for atmospheric studies.
 We have shown validation comparisons of ILAS version 5.2 data with observations from FIRS-2, SLS, and the JPL in situ UV photometer from a balloon flight on 30 April 1997 as part of the ADEOS Alaska Balloon Campaign. Back trajectories from the locations of the balloon data to the time of the nearest ILAS occultation show good spatial coincidences. As a result, we believe that the balloon and ILAS observations are covering air masses which have insignificant variability, making for robust comparisons. Mixing ratios of all the molecules discussed here (O3, H2O, N2O, HNO3, and NO2) from the ILAS and balloon observations show agreement to within the uncertainties.
 The usefulness of the validation exercise shown here is to highlight the robustness of the comparisons for all the molecules simultaneously, which tests for possible instrumental artifacts that will affect all molecules similarly, like pointing errors, or retrieval artifacts which could affect the mixing ratios of different molecules inversely, such as correlations between molecules. The absence of any such effects suggests that such systematic uncertainties are not significant relative to the random errors of the data.
 A number of the companion papers in this issue show thorough molecule specific comparisons for O3 [Sugita et al., 2002], NO2 and HNO3 [Irie et al., 2002], N2O and CH4 and H2O [Kanzawa et al., 2002]. These manuscripts test the robustness of the retrievals for a molecule over a range time, space, and validation instruments. All these studies show comparisons similar to what is shown here, that the ILAS data is in reasonable agreement with the validation data sets for most of the molecules. Since the different validation data sets occur over the life span of the ILAS instrument, it seems that the current version of ILAS data (Version 5.2) sufficiently accounts for any drifts in the data. Previous versions of the data had discrepancies with the validation data, which were most noticeable at the time of the FIRS-2 balloon flight.
 We are grateful to the National Space Development Agency of Japan (NASDA) for funding the balloon flight on 30 April 1997 from Fairbanks, Alaska as part of the Alaska Balloon Campaign correlative measurement program of the ADEOS satellite and to NASA's Upper Atmosphere Research program for support of work at SAO (grant NSG-5175) and JPL. We thank George Dobrowolski and George Laframboise of the National Research Council, Canada for their significant contribution to the development of the new beam splitter used to obtain the data presented here. We acknowledge the use of the NASA Goddard back trajectory calculations used in this study set up by Leslie Lait and Paul Newman. We are grateful to the Jet Propulsion Laboratory Atmospheric Ballooning group for gondola support and to the National Scientific Balloon Facility for launch services.