Vertical profiles of the gases O3, HNO3, NO2, N2O, H2O, and CH4, measured above Fairbanks, Alaska, during a May 1997 balloon flight of the JPL MkIV interferometer are compared with version 5.20 profiles measured by the ILAS instrument on board the ADEOS satellite. Both the ILAS and MkIV instruments employ the solar occultation technique, and for many of these gases, use the same absorption bands in their analyses. The results show good agreement for nearly all of these gases, after taking into account the slight differences in the potential vorticities of the sampled air masses and considering the error bars on the measurements.
 The Improved Limb Atmospheric Spectrometer (ILAS) measured the abundances of stratospheric O3, HNO3, NO2, N2O, H2O, and CH4 from their absorption of direct solar radiation in the mid-infrared [Sasano et al., 1999a]. ILAS was launched on board the Advanced Earth Observing Satellite (ADEOS) in August 1996. The ILAS experiment consisted of a Sun tracker, a telescope, and dual grating spectrometers. A 44 element pyroelectric detector array measured 850 to 1610 cm−1, while a 1024 element metal-oxide-semiconductor Si detector array covered the 12755 to 13280 cm−1 region containing the O2 A-band. Further details of the ILAS instrument are given by Suzuki et al. . ILAS successfully observed 5903 high-latitude occultations from November 1996 until the satellite failed in June 1997. This period saw the longest lasting Arctic winter vortex ever observed, with many Polar Stratospheric Cloud (PSC) encounters [Hayashida et al., 2000] and substantial ozone loss [Kreher et al., 1999; Sasano et al., 2000]. These high vertical resolution profile measurements have added substantially to our knowledge and understanding of chemistry and transport in the high-latitude stratosphere. However, it is still important to know to what confidence level the ILAS profiles (and their associated uncertainties) are realistic. Without this knowledge, artifacts in the ILAS data might be over-interpreted, or conversely, important features ignored. The National Space Development Agency of Japan (NASDA) and the Environmental Agency of the Japanese Government (EA) therefore sponsored several high-latitude balloon flights for the purpose of validating the instruments on board the ADEOS satellite. One of these balloon flights launched the MkIV interferometer from Fairbanks, Alaska, in May 1997.
 The Jet Propulsion Laboratory (JPL) MkIV Interferometer [Toon, 1991] is a high-resolution Fourier transform infrared (FTIR) spectrometer, designed to remotely sense the atmospheric composition. Optically, it is very similar to the Atmospheric Trace Molecule Spectroscopy (ATMOS) instrument [Farmer, 1987] which flew four times on the Space Shuttle. Like ATMOS and ILAS, the MkIV operates in solar occultation mode, in which the Sun is viewed through the limb of Earth's atmosphere. The MkIV instrument has performed 13 balloon flights over the past 12 years, several of which were dedicated to the validation of infrared instruments on board the NASA Upper Atmosphere Research Satellite (UARS). Several papers in the special section “Evaluation of the UARS Data” in Journal of Geophysical Research, 101(D6), 9539–10,473, 1996, made use of MkIV data.
 The MkIV retrievals were performed by a two-step procedure. First the spectra were fitted to determine the slant column abundances of the various gases. Then, the matrix equation relating these slant column abundances to the matrix of calculated slant path distances and the unknown volume mixing ratio (vmr) profile was solved, subject to a derivative constraint that favored the smoothest vmr profile that reasonably matched the measured slant column abundances. The MkIV retrievals are therefore not biased toward any particular a priori vmr profile; they are biased only to be smooth. More details on the MkIV profiles retrieval are given by Sen et al. .
 ILAS used an onion peeling retrieval method in which the entire spectrum was least squares fitted in order to simultaneously derive the vmr of each gas species for each layer. In the radiative transfer calculation to obtain theoretical ILAS signals, a look-up table was used for rapid calculation of infrared cross-sections as a function of air pressure and temperature. The look-up table was calculated in advance by the line-by-line method with the HITRAN96 line parameter database [Rothman et al., 1998]. Errors (internal error) in the retrieved gas profiles were estimated from residuals of the fitted spectrum and the Jacobi values for each gaseous species. Possible errors (external error) caused by other factors such as temperature, aerosol, and calibration uncertainties were modeled and evaluated in advance. An error bar provided as an ILAS product is a root sum square of the internal and external components. No constraint was applied in the retrieval procedure, and effects of the initial gas profiles on the retrieved results were negligibly small, in general. Details of the version 5.20 ILAS retrieval and error estimation procedures, which contain several improvements over earlier versions, are given by Yokota et al. .
 The MkIV balloon flight analyzed in this work was launched from Fairbanks on the evening of 7 May 1997. The 24 million cubic foot balloon (700,000 m3) attained a float altitude of 38 km during the night, and observations were made during sunrise on 8 May 1997, which was viewed to the NE of Fairbanks. Figure 1a shows the locations of the various MkIV tangent points (black squares) together with the locations of the ILAS occultations (colored symbols). Ten days of ILAS occultations were used in the comparison, all within 10° of longitude (∼500 km) of the MkIV profile. The early ones (blue) were closest in latitude to the MkIV profiles, whereas the later ones (red) were closest in time to the MkIV profiles.
 One of the advantages of balloonborne solar absorption spectrometry at high latitude is that the sunrise/sunset transitions take much longer than at lower latitudes. During the 8 May 1997 flight, the Sun took almost 2 hours to rise from a tangent altitude of 8 km to the balloon altitude of 38 km (as compared with 30 min at 34°N), during which time the MkIV instrument measured 39 limb spectra at an average tangent altitude separation of 0.8 km. In contrast, the ILAS occultations were each measured in just 10–20 s, owing to the 7 km/s orbital speed of the ADEOS satellite. This large difference in observation time, together with the poorer spectral resolution, explains why the random uncertainties of the ILAS profiles are substantially larger than those of the MkIV profiles.
 The circular MkIV field of view (FOV) subtends a diameter of 3.6 mrad, corresponding to 1.7 km at a tangent point 480 km distant from the balloon (typical for a 20 km tangent altitude). The ILAS field of view has a width of 1.57 km, and profiles are retrieved with a vertical resolution that decreases from about 3.2 km at 35 km, 2.6 km at 25 km, to about 1.9 km at 15 km altitude, owing to the effects of refraction. Thus the vertical resolutions of the ILAS and MkIV profiles are very similar.
 We had hoped to perform the balloon flight in mid-April, when the ILAS occultations would have been much closer in latitude to Fairbanks. However, unfavorable surface and stratospheric wind conditions delayed the launch by several weeks, during which time the ILAS occultations progressed southward. Fortunately, both the ILAS and the MkIV observations were made well outside the vortex, which was located over Europe and Russia during early May 1997 [Toon et al., 1999b, Plate 1]. The comparability of the air masses is confirmed in Figure 1b, which shows vertical profiles of modified potential vorticity (mPV) for the MkIV and the chosen ILAS occultations, on the basis of UKMO (UK Meteorological Office) analyses. Profiles of mPV were obtained by multiplying the actual potential vorticity by (PT/420)−9/2, as described by Lait , where PT is the potential temperature. This removes the exponential altitude dependence normally present in PV, facilitating comparison of PV values over a range of altitudes. The MkIV mPV profile is at the upper range of the spread in the ILAS values, as might be expected from its higher latitude. However, both the MkIV and ILAS mPV values (10−15 × 106 m2s−1kg−1K) are considerably smaller than typical vortex values (20−40 × 106 m2s−1kg−1K), confirming that the MkIV and ILAS observations were made in similar extra-vortex air masses.
3. Profile Comparisons
Figure 2 shows comparisons of the MkIV and ILAS vertical profiles for various atmospheric gases. In each case, the MkIV data are denoted by the black squares connected by a black line, and the ILAS profiles by the various colored symbols. MkIV measurement precisions are denoted by the error bars passing through the black squares. The range of ILAS uncertainties (total error) are defined by the dotted lines, color-coded to match the symbols. Although geometrical altitude has been used as the vertical ordinate in all of the subsequent comparison plots, we also tried using potential temperature (not shown), which should be conserved during adiabatic transport. But we found that this made little difference to the appearance of the biases between the two instruments, implying that the biases are not artifacts induced by vertical transport.
Figure 2a shows a comparison of the twelve closest ILAS O3 profiles with the one measured by MkIV. At the upper altitudes (>25 km) the ILAS data exhibit a decreasing trend over the 10-day period shown. Agreement is best for the later ILAS profiles (orange and red points), which were measured closest in time to the MkIV observations, although they are still 5–10% larger than those measured by MkIV outside the 18–22 km altitude range. This bias exceeds the combined error bars over the 24–34 km altitude range. This disagreement is unlikely to be an error in the MkIV profile, which agrees very well (better than 5%), with an ozone sonde launched from Fairbanks, Alaska, on the same day [Toon et al., 1999b, Plate 4]. It is also unlikely to be due to transport since the tracers N2O and CH4 do not exhibit trends. In a more comprehensive O3 comparison, Sugita et al.  found that the ILAS O3 data exhibited a high bias below 15 km, but that in general the ILAS O3 values were within 10% of the various correlative measurements. So a more likely explanation for the discrepancy between the MkIV and ILAS O3 profiles is that there were real spatial and temporal variations in O3 due to the fact that it was being photo-chemically destroyed during this time period. This is supported by the ILAS data themselves and by the decreasing trend of ground-based O3 columns measured from Fairbanks during this period [e.g., Toon et al., 1999a, Plate 1a].
Figure 2b shows a comparison of ILAS and MkIV profiles of HNO3. The agreement is good (better than 10%), except for altitudes between 25–31 km where all the ILAS profiles are biased 15% low, an amount that exceeds the combined error bars at 28–29 km altitude. This behavior confirms the conclusions of Koike et al.  who compared version 3.10 ILAS HNO3 with profiles measured by several different balloon instruments (the ILAS HNO3 profiles used in this work did not change significantly between versions 3.10 and 5.20). Around 15 km altitude the ILAS HNO3 profile of 970502 (left-pointing triangles) is substantially larger than the MkIV values (and the other ILAS profiles) by more than the combined error bars. However, this disagreement is clearly related to the much larger PV values (Figure 1b) at around 15 km altitude on this occasion. Conversely, the small values of ILAS HNO3 on 970505 (upward pointing stars) at 14–15 km altitude are related to low PV values.
Figure 2c shows a comparison of ILAS and MkIV profiles of NO2. Since the ILAS profiles were measured at sunset whereas the MkIV profile was measured at sunrise, it is not possible to directly compare them, due to the diurnal variation of NO2. However, we used a photochemical model calculation (R. J. Salawitch, private communication, 2000) based on the simultaneous MkIV measurements of NO, O3, ClNO3, N2O5, NOy, and Cly, to predict the diurnal variation of NO2. This way, we were able to calculate a sunset NO2 profile (unfilled squares) from the measured sunrise MkIV NO2 profile (filled squares). The former is larger, mainly due to daytime photolysis of N2O5. Previous MkIV balloon flights, in which sunrise and sunset profiles were obtained in the same air mass, confirm the correct diurnal behavior of the model [e.g., Sen et al., 1998]. The calculated sunset MkIV NO2 profile agrees well (better than 10%) with the ILAS profiles for altitudes above 16 km altitude.
Figure 2d compares ILAS and MkIV profiles of N2O. Between 18 and 25 km altitude agreement is very good (better than 5%). However, at higher altitudes all the ILAS profiles become larger, and at lower altitudes smaller, than the MkIV profile. These biases are typically smaller than the combined error bars, and so should not be over-interpreted. However, they are much larger than the scatter in the ILAS points, which is quite small except for the high PV points (left-pointing triangles) from 970505 around 15 km altitude.
Figure 2e compares ILAS and MkIV profiles of H2O. The ILAS values are biased systematically low: by over 15% between 18 and 22 km altitude, where the bias exceeds the combined error bars. At higher altitudes (above 35 km) the bias is much smaller. These conclusions are in broad agreement with those of Kanzawa et al. . The fact that the H2O bias between 22 and 35 km altitude is almost exactly equal to the ILAS uncertainties is purely coincidental.
Figure 2f compares ILAS and MkIV profiles of CH4. The ILAS values are biased high by up to 15% with respect to those measured by MkIV over the 17–29 km altitude range. However, this bias does not exceed the combined measurement uncertainties.
Figure 2g compares ILAS and MkIV profiles of the tracer H2O + 2*CH4, which should be approximately constant in the middle and upper stratosphere due to the fact that H2O is the predominant product of CH4 oxidation [e.g., Abbas et al., 1996]. The MkIV data indicates a value of 7.5 ± 0.1 above 20 km altitude. The ILAS value increases from about 7.2 ppm at 20 km to 7.4 ppm at 35 km altitude. The measurements agree to within their combined uncertainties. It is not clear whether the good agreement of the H2O + 2*CH4, relative to the poor agreement of the individual H2O and CH4 profiles, is fortuitous, or an indication of air mass differences between the MkIV and ILAS observations.
 In this work we have taken a slightly different approach to the validation effort. Instead of comparing ILAS profiles of a particular gas with correlative profiles measured by several different instruments [e.g., Sasano et al., 1999b; Lee et al., 1999; Sasano et al., 1999c; Koike et al., 2000], we attempt to compare all of the gases measured by ILAS to a single balloon measurement. This approach has the possible advantage that the similarities (or differences) in the biases between different species might provide insights as to their origin for future researchers. Another difference in our validation approach is that we have considered several ILAS occultations for comparison, not just the single ILAS occultation that was physically closest to the correlative measurement on the same day. Using several ILAS observations has two advantages: (1) it identifies any temporal trends in the ILAS data that might otherwise be misinterpreted as biases, and (2) it shows the level of consistency/scatter of the ILAS observations.
 The MkIV profiles used in this study have already been extensively compared with measurements taken by instruments on board the NASA ER-2 aircraft, which made several flights from Fairbanks during this same time period as part of the POLARIS campaign. The results of this comparison [Toon et al., 1999b] show good agreement for the species also considered in the ILAS comparison, even though the ER-2 measurements were, in general, a few degrees of latitude to the South of the MkIV balloon observations. In fact, the ILAS observations are probably better colocated with the ER-2 measurements than with those of MkIV. Unfortunately, the ER-2 measurements only extend up to 20 km altitude, and do not include HNO3. Otherwise, it might have made more sense to have directly compared the ILAS and ER-2 results, as done by Kanzawa et al.  for H2O.
 We consider it unlikely that the differences in the ILAS and MkIV profiles of the long-lived gases are the result of the temporal and geographical separation of the measurements. The ILAS profiles themselves show little trend with time or latitude (apart from a slight decrease of O3) indicating that they were made in a fairly homogenous air mass. This is confirmed by maps of PV, that show only slight gradients over Alaska during early May 1997 [Toon et al., 1999b, Plate 1], and by vertical profiles of PV (Figure 1b) which show typical midlatitude values with few exceptions, as discussed earlier in the Results section. Furthermore, ground-based observations made from Fairbanks by the MkIV instrument (while waiting for the balloon launch) show very little variation (less than 5%) in HF, a long-lived tracer whose column abundance has a strong poleward gradient [Toon et al., 1999a, Plate 1b].
 We speculate that it is more likely that the biases arise from the mismatch in the spectral resolutions of the two instruments. The ILAS resolution, which degrades from 9.3 cm−1 at 850 cm−1 to 33.5 cm−1 at 1610 cm−1, is substantially worse than that of the MkIV (0.01 cm−1). So although the MkIV and ILAS measurements covered the same spectral region and used many of the same absorption features, the much poorer spectral resolution of the ILAS measurements makes it more difficult to ascertain or verify the 0% and 100% signal levels from the spectra themselves since there are so few places in the low resolution spectrum where the limb transmittance is close to zero or unity. Furthermore, the high-resolution approach can much more easily distinguish narrow gas absorptions from underlying continuum absorptions (e.g., aerosol extinction), making the retrieved profiles less sensitive to the presence of aerosol. Finally, the high-resolution approach is also less sensitive to errors in the spectroscopy of interfering, gaseous absorption lines. This is because at high resolution such errors can be avoided by discarding portions of the spectrum that cannot be fitted adequately, but at low resolution one does not have this luxury since all parts of the spectrum must be used.
 Apart from a few exceptions (e.g., H2O), the version 5.20 ILAS profiles agree with those measured by the MkIV balloon interferometer to within their quoted uncertainties. For some gases (e.g., NO2, N2O, CH4) the ILAS uncertainties (total error) are much larger than the scatter in the various ILAS profiles or the differences between the ILAS and MkIV profiles, suggesting that for these gases the version 5.20 ILAS uncertainties are perhaps too conservative (large). In a more comprehensive study, Yokota et al.  shows that the ILAS measurement repeatability (scatter) was typically much smaller than the error bars calculated from the fit residuals, and speculated that systematic errors in the simulated spectra might be increasing these error estimates.
 The authors express their sincere thanks to the ILAS Project and Science Team members for their excellent work in developing the ILAS instrument, data processing system, in conducting validation experiments and analyses, and in improving the data processing algorithm. We also appreciate Richard Swinbank for supplying UKMO stratospheric assimilation data and related software. The data were processed at the ILAS Data Handling Facility, National Institute for Environmental Studies (NIES). The authors also express their gratitude to the National Scientific Balloon Facility, who conducted the MkIV balloon flight, and to Jim Riccio, Ron Howe, and Dave Petterson of JPL, who supported the field campaign. We also thank Ross Salawitch for use of his photochemical model in predicting sunset NO2 profiles from the measured sunrise profiles. This work was funded in part by the NASA Upper Atmosphere Research Program, with travel support from Japan's NASDA for the balloon flight, as part of the ADEOS Validation Campaign. A portion of this work was performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA.