Stratospheric and upper tropospheric distributions of peroxynitric acid (HO2NO2) were retrieved from limb infrared spectral measurements of the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on board the Environmental Satellite (ENVISAT). Single-profile precisions are estimated at 6–14 parts per trillion by volume (pptv) in the altitude range 7–17 km and 19–34 pptv from 17 to 42 km. The vertical resolution is 5 km in the upper tropospheric and stratospheric maxima. Highest stratospheric volume mixing ratios (VMRs) reaching 310 pptv at 27 km are observed at solstice conditions in subtropical latitudes and midlatitudes at the nighttime summer hemisphere, while lowest stratospheric peak VMRs as low as 38 pptv are found during polar winter near the pole. A second maximum in the upper troposphere and lower stratosphere appears from spring to the end of summer with maximum values of 80 pptv between 7 and 14 km. Retrievals based on spectroscopic line list data instead of absorption cross sections produce HO2NO2 distributions smaller by a factor of 1.5, on average. Earlier HO2NO2 measurements from balloon instruments are in good general agreement with the presented data set if the same spectroscopic data are used. Comparisons of MIPAS HO2NO2 distributions to results of the fifth-generation European Centre Hamburg general circulation model/Modular Earth Submodel System 1 (ECHAM5/MESSy1) provide agreement within 20% if near-infrared photolysis is considered. With the newly available tabulated absorption cross sections and the improved photolysis modeling, former discrepancies between HO2NO2 observations and model calculations can be considered to be largely resolved.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 HO2NO2 is a minor constituent of the stratosphere. Although its concentration is about an order of magnitude smaller than other compounds of the NOy family, it is important for understanding the stratospheric chemistry as it couples the NOx and HOx catalytic cycles of ozone destruction. It is formed by the reaction
 Thermal decomposition (R3) is fast, leading to short lifetimes of HO2NO2 (≤1 day) at low altitudes, below 7 km. The illumination dependence of sink reactions R2a and R2b (directly via hv) as well as R4 (via the diurnal cycle of OH) explain the diurnal cycle of HO2NO2, which is characterized by daytime reduction. The sequence (R1 + R4) describes the NOx-catalyzed loss of HOx, and represents a significant sink of odd-hydrogen radicals in the stratosphere and at higher NOx levels in the upper troposphere [Roehl et al., 2002; Salawitch et al., 2002; Brasseur and Solomon, 2005]. HOx radicals in the lower stratosphere are the dominant sink for photochemical loss of ozone, and ozone production in the upper troposphere proceeds via coupled HOx:NOx photochemistry, limited because of loss of HOx via reactions R1 + R4 [Salawitch et al., 2002]. For this reason, global HO2NO2 distributions contribute to quantify sources and sinks of HOx and, in turn, constrain the ozone photochemistry of the upper troposphere and lower stratosphere.
 From model calculations, latitudinal, seasonal and small diurnal variations of the HO2NO2 abundance and a dependence on the solar zenith angle are expected [e.g., Evans et al., 2003; R. Ruhnke, personal communication, 2004]. A stratospheric volume mixing ratio (VMR) maximum is expected during springtime in midlatitudes and at the summer poles between 20 and 30 km, depending on season. Minima are expected at the tropics and in polar winter. Nighttime maximum VMRs are calculated to be up to a factor of 2 larger than daytime values above 10 hPa. Further, a second maximum at high latitudes during springtime is expected in the upper troposphere/lowermost stratosphere [Salawitch et al., 2002].
 Photolysis of HO2NO2 ((R2a) and (R2b)) in the ultraviolet (UV) is well established. However, Donaldson et al.  suggested that photolysis in the near infrared (NIR) may also be important, and Wennberg  concluded that this process could be an important source of HOx at high solar zenith angles. Increased levels of HOx, in turn, would have direct impact on upper tropospheric (UT) and lower stratospheric (LS) ozone abundances. Increased HOx would further alter the coupling of the NOx and HOx catalytic ozone destruction cycles, and have impact on NOx concentrations in the stratosphere. The NIR photolysis path is expected to be more important for the high-latitude spring [Salawitch et al., 2002], since, through strong UV attenuation by high columns of ozone, UV photolysis is less efficient, but the long days favor photolytic sinks. Evans et al.  showed that according to their model simulations the NIR photolysis is most effective in reducing HO2NO2 at altitudes around 18 km, with maximum reduction of up to 95% during polar winter.
 New absorption cross sections of HO2NO2 photolysis in the NIR were reported by Roehl et al. . The inclusion of NIR photolysis in chemical models led to better agreement with observations [Salawitch et al., 2002; Evans et al., 2003]. However, significant differences still remained, and the comparison suffered from the sparse HO2NO2 data set available in the past.
 Observations of atmospheric HO2NO2 profiles are available from balloon-borne and spaceborne Fourier transform infrared (FTIR) instruments. Rinsland et al.  first detected HO2NO2 absorptions in Atmospheric Trace Molecule Spectroscopy (ATMOS)/Spacelab-1 spectra recorded in 1985. Self-consistent HO2NO2 ATMOS data sets from the Spacelab-1 and the 3 ATLAS missions were provided by Irion et al. . Goldman et al.  improved HO2NO2 spectroscopic parameters from ground-based and balloon-borne solar absorption spectra. A series of balloon flights provided HO2NO2 observations in the Northern high and middle latitudes both from the MkIV solar occultation instrument [Sen et al., 1998; Osterman et al., 1999; Salawitch et al., 2002] and from the Michelson Interferometer for Passive Atmospheric Sounding B2 (MIPAS-B2) balloon-borne limb emission FTIR spectrometer [Wetzel et al., 1997, 2002; Stowasser et al., 2002; Evans et al., 2003].
 In this paper we describe the retrieval approach used to derive global HO2NO2 distributions from MIPAS spectral data, discuss the HO2NO2 data in terms of precision, accuracy, and vertical resolution, and present the global data set (pole to pole, day and night) derived for all seasons. The sensitivity of the resulting global distributions to two different spectroscopic input data sets has been quantified, and latitude- and altitude-dependent transformation factors have been derived. In addition, a comparison to existing HO2NO2 profiles from balloon-borne instruments has been performed, in order to demonstrate the consistency of this new data set with former measurements if the same spectroscopic data are used. Finally, global distributions of HO2NO2 are compared to results of the atmospheric chemistry general circulation model European Centre Hamburg general circulation model/Modular Earth Submodel System 1 (ECHAM5/MESSy1) [Jöckel et al., 2006] with and without consideration of the NIR photolysis channel.
2. MIPAS Data
 MIPAS (Michelson Interferometer for Passive Atmospheric Sounding) is a high-resolution midinfrared limb emission sounder onboard ENVISAT (Environmental Satellite) designed for measurement of trace species in the lower to middle atmosphere [European Space Agency, 2000; Fischer and Oelhaf, 1996]. It is one of the three atmospheric chemistry instruments onboard ENVISAT which was launched on 1 March 2002 into a Sun-synchronous polar orbit at about 800 km altitude with ≈1000 and 2200 LT local equatorial crossing time, and an inclination of 98.55°. This orbital geometry results in a near to complete coverage of the globe where, in particular, the poles can be sounded during day and (polar) night.
 MIPAS became operational in July 2002 and provided around 1000 vertical profiles of the infrared atmospheric spectral radiance along 14.3 orbits per day with full spectral resolution of 0.035 cm−1 until late March 2004, when the spectrally high-resolved measurements had to be suspended because of instrument failure. In its standard observation mode MIPAS covered an altitude range of 6 to 68 km with vertical scanning step width of 3 km up to 42 km, and above at 47 km, 52, 60 and 68 km. The instantaneous field of view (FOV) is 30 km in the horizontal and about 3 km in the vertical direction [Endemann et al., 1996]. Data presented here were recorded from September 2002 to March 2004. The data analysis presented here is based on the ESA (European Space Agency) provided level 1b data product (versions IPF4.61/4.62) which are geolocated, phase-corrected, calibrated radiance spectra [Nett et al., 1999]. Our data analysis is independent of the level 2 data analysis of ESA whose operational data product covers vertical profiles of temperature, pressure, O3, H2O, HNO3, CH4, N2O, and NO2 [Ridolfi et al., 2000] and provides, besides another version of the species covered by the operational data product, a set of species complementary to that.
3. Retrieval of HO2NO2
3.1. Retrieval Strategy
 The HO2NO2 distributions discussed in this paper were retrieved with a scientific data processing system developed at IMK (Institut für Meteorologie und Klimaforschung) together with IAA (Instituto de Astrofísica de Andalucía, Granada, Spain) [von Clarmann et al., 2003b]. HO2NO2 vertical profiles were retrieved from the 802 to 804 cm−1 spectral region in MIPAS channel A which contains the Q branch of the ν6 band at 802.7 cm−1, the only signature of HO2NO2 in the midinfrared for which spectroscopic data are publicly available. The retrieval method applied is an iterative nonlinear multiparameter least squares fit with a smoothing constraint. Details on the implementation are given by von Clarmann et al. [2003a]. The IMK/IAA processor uses the Karlsruhe Optimized and Precise Radiative Transfer Algorithm (KOPRA) [Stiller et al., 2002] for simulation of the radiative transfer.
 Since the retrieval is performed on a vertical grid of 1 km (below 44 km) to 2 km (above 44 km) step width, which is finer than the tangent altitude grid, stabilization of the retrieval is necessary. This is done by Tikhonov-type regularization [Tikhonov, 1963] as implemented by Steck and von Clarmann  and Steck , which minimizes the first-order finite difference quotients Δ VMR/Δz (with z = altitude) of adjacent altitudes. An all-zero profile has been used as a priori profile, to avoid introducing any artificial structures into the resulting profiles.
 Prior to the HO2NO2 retrieval, retrieval of a residual frequency shift, simultaneous temperature and line-of-sight pointing information [von Clarmann et al., 2003a] and the retrieval of O3, H2O, HNO3, CH4, N2O, N2O5, ClONO2, ClO, and CFC-11 have been performed. These results are used, unless otherwise noted, as a priori information on the profiles of “interfering” species during the HO2NO2 retrieval to minimize error propagation due to their VMR uncertainties. In particular, temperature and line-of-sight pointing information as retrieved from the spectra is used in the HO2NO2 retrievals.
 Simultaneously with HO2NO2, the vertical profiles of O3, CFC-113, and ClONO2, as well as an altitude-independent radiometric zero level correction and, below 32 km altitude, an empirical background emission continuum profile have been fitted. In the case of O3 and ClONO2, the simultaneous joint retrieval has been performed in order to correct for potential inconsistencies in spectroscopic data of these species and to keep error propagation caused by spectroscopic uncertainties small. While for H2O, HNO3, and ClO preretrieved profiles have been used to account for their signatures in the microwindow, climatological abundance profiles [Kiefer et al., 2002] have been used for further interfering species, i.e., CO2, NO2, NH3, OCS, HCN, CH3Cl, C2H2, C2H6, C2H4, COF2, CCl4, HCFC-22, acetone, CH3CCl3, peroxyacetyl nitrate (PAN), and CH3OH. This is justified because for some of these species the climatological variability is small, while for others their spectral contribution is too low to trigger any substantial error propagation even if the actual atmosphere deviates from the climatology.
 Local spherical homogeneity of the atmosphere has been assumed; that is, the atmospheric state parameters related to one limb sounding sequence are allowed to vary with altitude only. An example of the spectral fit of the HO2NO2 signature is given in Figure 1.
3.2. Spectroscopic Data
 Spectroscopic data for all interfering species were taken from a dedicated MIPAS spectroscopy database [Flaud et al., 2003] which is an upgrade to the High-Resolution Transmission Molecular Absorption Database (HITRAN) year 2000 edition [Rothman et al., 2003]. For HO2NO2, two data sets are available. Most former measurements relied on the spectroscopic line parameters as provided by Brown et al.  for the ATMOS experiment which is a pseudo line list with constant ad hoc lower-state energies and line widths rather than true rovibrational line data, hence not expected to describe the temperature dependence of the HO2NO2 Q branch correctly [Rinsland et al., 1986]. Further, as May and Friedl  point out, not all fundamental vibrational frequencies of HO2NO2 have been observed which hinders correct assessment of the vibrational partition function. The spectroscopic line list data were used for all our data versions up to V3O_HNO4_9 (inclusive) (hereinafter referred to as version 9). More recently, a data set of tabulated absorption cross sections for HO2NO2 measured at 220 K and 0.05 torr for the spectral range 780 to 830 cm−1 has been made available [May and Friedl, 1993; Rothman et al., 2005; Brown et al., 1996]. These data were used for our otherwise identical data version V3O_HNO4_10 (hereinafter referred to as version 10). Both data versions 9 and 10 are based on ESA offline reprocessing of the level 1b data set covering July 2002 to late March 2004 (versions IPF4.61/4.62).
 The fit quality in terms of χ2 is comparable for both spectroscopic data sets. However, resulting version 10 abundances are higher by a nearly altitude-constant factor of 1.5 (see Table 1). A part of this bias is caused by the different band intensities of both spectroscopic data sets, which accounts for about 15 to 20% difference. Even more important is the fact that the line list contains transitions from 802.48 to 803.65 cm−1 only, while the tabulated cross-section data also includes a considerable spectrally flat HO2NO2 signal below 802.48 cm−1 and above 803.65 cm−1 (see Figure 2). When the line list is used, the retrieval assigns this flat portion of the HO2NO2 signal to the background continuum (see von Clarmann et al. [2003a] for the rationale of background continuum fitting). Since the background continuum radiance is forced to be constant over the entire window 802.0–804.0 cm−1, a considerable fraction of the HO2NO2 Q branch emission is taken as continuum emission, and, in consequence, retrieved HO2NO2 mixing ratios underestimate the real atmospheric abundances. Therefore we conclude that version 10 retrievals are more reliable. The global HO2NO2 data set presented throughout this paper, except for the comparison to balloon-borne observations described in section 5, is based on the absorption cross sections by May and Friedl , which refers to the higher HO2NO2 VMRs.
Table 1. Altitude- and Latitude-Dependent Scaling Factors Between HO2NO2 Profiles Retrieved on the Basis of Absorption Cross-Section Data (Data Version 10) and Spectroscopic Line List Data (Data Version 9)
The scaling factors were derived from a number of test cases covering various seasons and all latitudinal zones (8 orbits, 575 profiles). The absorption cross sections provide the higher VMRs. Scaling factors together with their uncertainties (ensemble standard deviations divided by square root of the number of the ensemble members) are given. Missing entries are caused by missing profile data, due to clouds or insufficient sensitivity of the measurement.
The global average is derived from all profiles available and is thus a number-of-profiles weighted average of the values provided for the latitude bands.
The HO2NO2 VMRs of both profiles are virtually zero, which makes a scaling factor meaningless; for this reason, no entry is given. NA means not applicable.
 The only drawback of the tabulated absorption cross sections by May and Friedl  is that no information on their temperature dependence is available. The data are provided for one single temperature, namely 220 K. To assess the differences in HO2NO2 data versions 9 and 10 due to neglect of the temperature dependence of the tabulated absorption cross sections a sensitivity study has been performed. For one test orbit, profiles were retrieved using the line data set, with consideration of temperature dependence as a reference run, and deactivated temperature dependence, with temperature fixed at 220 K, as test run. Reference absorption cross sections calculated from the spectroscopic line list data vary by ±40% for the stratospheric/upper tropospheric temperature range relative to those at 220 K. This translates into mean differences of retrieved profiles (averaged over a test orbit) of up to 17 parts per trillion by volume (pptv) (40%) above the HO2NO2 VMR maximum, 7 pptv (6%) in the VMR peak, and less than 5 pptv (up to 30%) below the stratospheric VMR peak (see Figure 3). Profiles of the test run are always lower than the reference run above and in the VMR peak, and higher below, reflecting the temperature distribution in the stratosphere. For individual profiles, differences of 50% or even more can occur, for example for the very low temperatures during the Antarctic polar winter.
 However, the temperature dependence of the absorption cross sections calculated from the line list data is not necessarily correct, as explicitly stated by Rinsland et al. . We therefore conclude that the temperature dependence of the Q branch absorption is not correctly modelled by either of the two spectroscopic data sets; the HO2NO2 continuum absorption apart of the Q branch, however, can only be accounted for by the tabulated absorption cross sections. For this reason we consider the version 10 of HO2NO2, based on the tabulated cross-section spectroscopic data, as the best currently available. On the other hand, as a continuation of their initial work, Friedl et al.  derived a line list for the Q branch region consisting of 28000 lines, which allows to correctly model temperature and pressure dependence. These most recent spectroscopic data, however, have not been included in the HITRAN database, probably because such a high number of lines within a small spectral interval is difficult to handle in atmospheric radiative transfer calculations. Further improvement will be possible by providing new spectroscopic data as, for example, pressure-temperature-dependent tabulated absorption cross sections.
 When we compare MIPAS data to measurements from other instruments, data version 9 is used for reasons of consistency: Analysis of both MkIV data (G. Toon and R. Salawitch, personal communication, 2006) and MIPAS-B data [Stowasser et al., 2002; Wetzel et al., 1997, 2002] was based on the line list data, and the same spectral range was used. Unfortunately, validation with coincident measurements will not provide more insight into spectroscopic data, because all HO2NO2 measurements published by now have been made in the same spectral region and depend on the same spectroscopic data.
 The correction factor to convert HO2NO2 profiles from one version to another is nearly independent on latitude and altitude (see Table 1) with cross-section results being higher than line list results. It has some variation mostly in the upper troposphere and lower stratosphere. The scaling factors in Table 1 can be used to transform the profiles retrieved on basis of spectroscopic line data with good accuracy.
3.3. Error Assessment
 Two representative profiles together with the estimated standard deviation (ESD, error due to measurement noise) are shown in Figure 4. The strength of the regularization, being altitude-dependent in our approach [Steck, 2002], has been adjusted such that the retrieved profile represents about 6 degrees of freedom (trace of the averaging kernel matrix A, see Rodgers  for the theory behind this approach), corresponding to an altitude resolution of 4 to 6 km in the upper troposphere and stratosphere. This vertical resolution allows to resolve the maxima in the HO2NO2 VMR profiles, expected to occur in the middle stratosphere and the upper troposphere. The rows and columns of the averaging kernel matrix A from which the vertical resolution can be derived, e.g., as full width at half maximum of the rows, are shown in Figure 5.
Table 2 summarizes the uncertainties due to all known and quantifiable error sources for a Northern polar summer profile (measured at 85.2° N and 146.8°E on 24 June 2003, 0054 UTC, at 62.1° solar zenith angle). Error budget calculations for 140 different profiles representing various latitude zones and seasons have confirmed that the absolute errors vary little, except for the polar winter region where polar stratospheric clouds (PSCs) are present. The error budget is dominated by the measurement noise error (ESD) which is around 26 pptv at 34 km, 32 pptv at 27 km, and 7 pptv at the tropospheric maximum. Further error sources of minor relevance are temperature and line-of-sight pointing uncertainties (obtained from preceding retrievals) which contribute by 5 and 9 pptv, respectively, at most, to the total error at altitudes around 30 km. The residual uncertainty of the instrumental line shape provides a systematic error contribution of 8 pptv above 34 km. Uncertainties of neither interfering species nor any other parameters contribute to the total error budget to any substantial amount. The total precision of the HO2NO2 profile, i.e., its total random error including measurement noise and parameter uncertainties with random characteristics in the time domain, is estimated at 19 to 34 pptv (or 11 to 26%) of the midstratosphere maximum. The total precision of the tropospheric maximum is about 9 pptv or 25%. The uncertainty of the HO2NO2 tabulated absorption cross sections spectroscopic line list parameters is estimated at 15% [May and Friedl, 1993], however, not including uncertainties due to unknown temperature and pressure dependence. Thus the overall accuracy of individual HO2NO2 profiles is estimated at 20% to 30% of the maximum measured abundances at the stratospheric peak.
Table 2. Total Error Budget for Data Version 10 With Percentage Errors in Parenthesesa
For version 9, the percentage errors apply, while absolute errors have to be divided by the scaling factors reported in Table 1.
MIPAS NESR is on average 14 nW/(sr cm2 cm−1) (for apodized spectra) in the HO2NO2 microwindow.
Temperature uncertainty was assumed to be 1 K.
Residual line-of-sight pointing uncertainty was assumed to be 0.15 km.
The quadratic sum of uncertainties due to the following error sources is given here: Interfering nonjointly fitted species VMRs, neglect of horizontal inhomogeneity in temperature, residual radiance calibration uncertainty, residual frequency calibration uncertainty, and forward modeling errors, in particular, neglect of nonlocal thermodynamic equilibrium effects.
Residual instrumental line shape uncertainty was assumed to be 3%.
This column presents the quadratic sum of all errors reported in the second through fifth columns.
This column presents the quadratic sum of all errors reported in the sixth and seventh columns and the spectroscopic uncertainty, estimated at 15%, according to the uncertainty given for the cross sections [May and Friedl, 1993] and assuming linear error propagation.
The VMR profile is shown in Figure 4, bottom plot. Percentage errors in this table refer to this profile.
The empirical standard deviation for the zonal mean ensemble related to June 2003, 82.5°N–90°N, is given for comparison with the total precision.
 For observations within the polar winter vortex, the total precision of HO2NO2 is often significantly worse than that given in Table 2 as representative for other regions/periods. This is due to the occurrence of PSCs in the lower stratosphere up to 25 km or higher which do not allow HO2NO2 VMRs to be retrieved at these altitudes from cloud-contaminated spectra. This drawback results in severely increased retrieval uncertainties in terms of estimated standard deviations (ESDs) and reduced degrees of freedom for the remaining parts of the profiles.
 The estimated error budget is compared to the empirical standard deviation of day and night zonal mean values, retrieved for December 2002, and March, June, and September 2003 (see Figure 6). Zonal mean values of the HO2NO2 VMR have been calculated by averaging over all available data of one month, separated into day and night measurements, within 5° latitude bins. For all cases, the standard deviations are nearly constant over wide latitudinal regions, with values of 25 to 50 pptv between 25 and 37 km, which confirms our error estimates and implies low natural variability within the observed scenes. Some higher standard deviations with values up to 80 pptv are found in the polar regions. These are, in general, attributed to enhanced atmospheric variability related to solar zenith angle variation as well as chemical and dynamical processes in both vortices which considerably altered their NOy abundances (for more details, see Funke et al.  and Konopka et al. ). It should be noted that in December 2002 and June 2003, the observed standard deviations poleward of ≈70° N and S, respectively, reflect the variability under conditions of very short days or total polar night. Maxima in standard deviations far beyond 80 pptv as observed during polar night in December and June are, at one hand, attributed to ensembles of profiles with high natural variability, including very low profiles. In particular, the Arctic vortex usually reveals high variability with respect to its latitude coverage while we have performed zonal averaging over geographical latitudes; that is, most probably, vortex and nonvortex air has been averaged in this case (December 2002) as well as in the comparisons to MIPAS-B2 profiles. For June 2003 (Antarctic winter) previous studies showed that high NOx abundances of mesospheric/lower thermospheric origin were transported downward into the stratosphere [Funke et al., 2005] which altered the stratospheric NOy budget considerably over few weeks. Daily zonal mean values of HO2NO2 over the Antarctic were virtually zero on 6 and 8 June, revealed a maximum of up to 150 pptv around 27 km on 12 June and were back to about 75 pptv on 24 June. This sudden and short-lived HO2NO2 enhancement probably was due to the NOx intrusions from the upper atmosphere, and produced the high variability in the monthly zonal means. On the other hand, however, the large standard deviations also reflect the deteriorated total precision of polar winter vortex retrievals due to PSC occurrence.
 A second maximum of the standard deviations is found in tropics and northern midlatitudes around 10 km. Here the standard deviation reaches values of up to 40 pptv. The standard deviations are lower, with values below 25 pptv, in subpolar and polar regions where we expect the late spring/summer upper tropospheric/lower stratospheric maximum. This indicates that the climatological variability is smaller in this situation, and the estimated total precision is roughly consistent with the standard deviation of the data set. The reasonable agreement between the standard deviation of the data set with the theoretically estimated precision rules out that our error estimation is overoptimistic and confirms that all relevant randomly varying error sources have been considered.
4. Global Distribution of HO2NO2
 HO2NO2 profiles were retrieved from MIPAS spectra in the period September 2002 to late March 2004. From single HO2NO2 profiles, monthly zonal mean profiles were derived by averaging over all available profiles within 5° latitude bins. In Figure 7 examples for March and September 2003, representing equinox conditions, and June 2003 and December 2002, representing solstice conditions, are shown. The mean global distributions at equinox conditions have their maxima in terms of mixing ratios in the subtropics to midlatitudes at about 27 km with values up to 280 pptv, with a strong symmetry for the two hemispheres, while mean volume mixing ratio values over the poles reach 165 pptv in Arctic spring at 25 km altitude. The Antarctic spring seasons in the two years observed were very different: while in September 2002 (not shown) similar values as in Arctic spring (up to 165 pptv) were observed, maximum values in September 2003 did not exceed 80 pptv poleward of 67.5°S. We explain this by the unusual Antarctic vortex split and displacement occurring in September 2002, which, however, is the usual behavior of the Arctic polar vortex, although not accounted for in the zonal averaging. Averages over equivalent latitudes would have probably provided similar results for the various Arctic and Antarctic spring situations. At the autumn poles, the maximum VMRs occur at about 25 km with abundances of up to 120 pptv. At solstice conditions (December and June) the stratospheric maxima are shifted toward the summer hemisphere, reaching 310 pptv in June at 30° N during night, and 295 pptv in December in Southern subtropical latitudes at 27 km altitude. Polar day values over the summer pole reach 190 pptv at 25 km altitude, and maximum values in the winter polar night are as low as 45 pptv at 27 km in December 2002 and June 2003, and lower than 70 pptv in December 2003. A second maximum in the upper troposphere and lowermost stratosphere appears from spring to the end of summer with maximum values up to 80 pptv between 7 and 14 km which is more pronounced in the Northern Hemisphere. The uncertainties of the mean values in the stratospheric maximum are, dependent on ensemble size, around 10%, while in the upper tropospheric/lower stratospheric maximum, the uncertainties of the mean values are 3 to 20%.
 Day-night differences are most pronounced above the stratospheric maximum, at 30 to 40 km, with daytime values being typically 30 pptv lower (see Figure 8). This represents a diurnal amplitude of a factor of 2 at this height, in good agreement with model predictions [see, e.g., Brasseur and Solomon, 2005], while at lower altitudes, e.g., 20 km, a pronounced diurnal variation is neither expected nor observed. Pronounced maximum day-night differences occur in December in Southern and in June in Northern midlatitudes, and in April/May (not shown) and August/September for subpolar to tropical latitudes. Interestingly, during equinox at the autumn poles, poleward of 80°, the daytime VMRs are higher than the nighttime VMRs all over the observed stratospheric altitude range, with differences of 30 pptv and beyond. This, however, is understood as a sampling effect, since HO2NO2 decreases at the autumn poles during September/March, while illumination turns from polar day to polar night, and thus the daytime observations are obtained earlier in the month. Contrary to the stratospheric maximum, in the upper tropospheric/lower stratospheric maximum day-night differences occur only during April to June with maximum values of 20 pptv.
 These day-night differences, however, are not representative of the full amplitude of the diurnal cycle, but related to the nearly constant local measurement times (approximately 1000 and 2200 LT) of MIPAS. Thus our observations do not necessarily capture the diurnal minima and maxima, and are further affected by different illumination conditions (solar zenith angle, time since sunrise/sunset) at different latitudes.
5. Comparison to Other Measurements
 In order to confirm the reliability of the MIPAS/ENVISAT measurements, monthly zonal mean profiles for 5° latitude bins have been compared to balloon-borne HO2NO2 profiles. We have checked if there is any indication that the balloon-borne profiles were not part of the monthly ensemble observed by MIPAS. A detailed quantitative validation of the MIPAS HO2NO2 observations with coincident measurements will be presented in a subsequent publication. Two balloon-borne instruments have contributed so far to the current knowledge on atmospheric HO2NO2 distributions: the Fourier transform infrared (FTIR) solar occultation spectrometer MkIV [Toon, 1991], and the balloon-borne version of the MIPAS instrument, the FTIR limb emission spectrometer MIPAS-B2 [Friedl-Vallon et al., 2004].
 Two HO2NO2 profiles observed by the MkIV spectrometer have been published so far: one measured during a sunset occultation on 25 September 1993 at a balloon flight from Fort Sumner, New Mexico (35° N) [Salawitch et al., 2002; Sen et al., 1998], and another measured during a sunrise occultation on 8 May 1997 at a balloon flight from Fairbanks, Alaska at 65° to 70° N [Osterman et al., 1999; Salawitch et al., 2002]. Further, a profile measured during sunset during a balloon flight from Fort Sumner, New Mexico (35° N) on 19 September 2003 was made available to us (G. Toon and B. Sen, unpublished data and personal communication, 2006). In contrast to the occultation observations, MIPAS measures always at about 1000 and 2200 local time which is about 4 hours after sunrise and sunset, respectively, in September, and under polar daytime conditions for both observation times at 70° N in May. The MkIV profiles were retrieved on basis of the spectroscopic line list provided by Brown et al.  (G. Toon, R. Salawitch, personal communication, 2006), using the same spectral range as that used for MIPAS retrievals; further, according to Salawitch et al. [2002, Figure 1], any continuum-like spectral contribution from HO2NO2 apart from the immediate Q branch signature was not considered in their spectral analysis. Therefore, in order to be as consistent as possible, we compare MIPAS data version 9 to the MkIV profiles.
Figure 9, top plot, compares the MkIV profiles measured at 35° N in September at sunset with zonal mean profiles from MIPAS observations generated by averaging over all September 2002 and September 2003 profiles, respectively, within a 5° latitude bin around 35° N. The MIPAS mean profiles for September 2002 and 2003 differ little, by only 12 pptv, while day-night differences are 22 pptv at maximum (September 2003) in the peak altitude. The 1993 MkIV profile is higher by about 40 pptv than the MIPAS nighttime profiles in and above the profile maximum while the 2003 MkIV profile is higher by only 11 pptv at the peak and compares well above and below. All profiles are well within the combined uncertainties, given by the ensemble standard deviation in case of MIPAS profiles, and the total error in case of the MkIV profiles. The altitude of the maximum and the VMRs in the altitude range below agree very well.
Figure 9, bottom plot, compares the MkIV profile measured at 65° to 70° N in May 1997 at sunrise with zonal mean profiles of MIPAS for May 2003 at 70° and 65° N, respectively. The stratospheric parts of the profiles again compare well within the combined uncertainties, while for the tropospheric peak, a shift in altitude of about 2 km seems to be apparent, and the tropospheric peak in the MIPAS profiles is much smaller. This difference is most probably attributed to a different atmospheric situation, e.g., to different tropopause altitudes.
5.2. Balloon-Borne Version of the MIPAS Instrument, MIPAS-B2
 From MIPAS-B2, six HO2NO2 profiles have been published, covering nighttime polar winter conditions (January, February and March at 68°N), as well as nighttime midlatitude spring (April, 41°N), and nighttime midlatitude summer (July, 44°N) conditions. All MIPAS-B2 profiles are compared to zonal mean profiles from MIPAS/ENVISAT measurements, generated by averaging over all available nighttime profiles for the respective month and a 5° latitude bin around 65°N, 70°N, 45°N, or 40°N, respectively (see Figure 10). All MIPAS-B2 profiles were retrieved on basis of spectroscopic line data as provided by Brown et al. . Similar to the MIPAS/ENVISAT analysis, the 802 to 804 cm−1 spectral range was used, and an offset accounting for continuum contributions of unknown origin was simultaneously fitted (G. Wetzel, personal communication, 2006). For these reasons, we compare these data also to the MIPAS/ENVISAT data version 9.
 The top left plot of Figure 10 compares two MIPAS-B2 profiles measured on 27 January 1999 [Stowasser et al., 2002] and 11 January 2001 [Evans et al., 2003], respectively, at nighttime balloon flights from Kiruna, Sweden (68°N) with MIPAS/ENVISAT January 2003 nighttime zonal mean profiles for 65°N and 70°N. The natural variability of the individual MIPAS/ENVISAT profiles, included in the standard deviations of the zonal means, is very high, similar to December conditions as shown in Figure 6. This is also reflected by the very different MIPAS-B2 profiles from the two flights. Both MIPAS-B2 profiles are within the 1 σ standard deviation of the MIPAS/ENVISAT zonal means, indicating that there is no evidence for systematic deviations.
 For February conditions (top right plot of Figure 10), a MIPAS-B2 profile measured on 11 February 1995 during night from Kiruna, Sweden [Wetzel et al., 1997] is compared to February 2003 zonal mean profiles for 65°N and 70°N from MIPAS/ENVISAT. Although its peak is much narrower, the MIPAS-B2 profile is within the large 1 σ standard deviations of the MIPAS/ENVISAT zonal mean profile, and the differences are all within the combined error bars. For March conditions (middle left plot of Figure 10), the MIPAS-B2 profile (measured on 24 March 1997 from Kiruna [Wetzel et al., 2002]) again seems to be systematically lower than the MIPAS/ENVISAT zonal mean profile. However, the differences are still within the combined error bars. Midlatitude observations of MIPAS-B2 (middle right plot of Figure 10: profile measured on 30 April 1999 from Air sur l'Adour, France, 41°N [Evans et al., 2003]; bottom left plot of Figure 10: profile measured on 2 July 1997 from Air sur l'Adour, France [Wetzel et al., 2002]) agree very well with MIPAS/ENVISAT zonal mean profiles for April 2003 and July 2003, respectively.
 In summary, no evidence for systematic deviations from previous HO2NO2 observations by MIPAS-B2 has been found, although there seems to be a tendency for MIPAS/ENVISAT zonal mean profiles to be higher for the Northern high-latitude winter/spring stratosphere. Considering that no trends have been considered, zonal averaging has not accounted for the actual latitudinal coverage of the polar vortex, and comparisons suffer from some diurnal mismatch, the profiles agree well. A dedicated quantitative validation study with correlative measurements is under way.
6. Comparison to Model Calculations
 We compare our results with calculations performed with the new atmospheric chemistry general circulation model ECHAM5/MESSy (version 1.1), further denoted as E5/M1 [Jöckel et al., 2005, 2006]. E5/M1 has been applied in T42L90MA resolution (triangular truncation at wave number 42 for the spectral core of ECHAM5 [Röckner et al., 2006], 90 vertical layers up to ≈1 Pa [Giorgetta et al., 2006]).
 The prognostic variables (surface pressure, temperature, divergence, and vorticity) of ECHAM5 are nudged in the troposphere (the latter three between the 4th layer above the surface up to ≈100 hPa) to ECMWF analysis data. As shown by Jöckel et al.  this is sufficient to reproduce the observed tropospheric meteorology and the wave driven forcing of the stratosphere, and allows a point-to-point comparison of simulated data (such as temperature and constituents) with observations, as done for Figures 11 and 12.
 The comprehensive model setup comprises all relevant processes for global chemistry simulations from the troposphere to the mesosphere: in particular the submodel MECCA [Sander et al., 2005] for gas phase chemistry, the submodel PSC [Buchholz, 2005] for the physical and chemical effects of polar stratospheric clouds, the submodel HETCHEM for stratospheric sulfate aerosol effects, and the submodel JVAL (based on work by Landgraf and Crutzen ) for the online calculation of photolysis rates. In JVAL we optionally included the near-IR photolysis of HO2NO2 [Roehl et al., 2002] and repeated the simulation S1 from 1998 to 2005 in the nudged mode out of Jöckel et al.  for September 2002. Details on emission and removal processes are described by Jöckel et al.  and references therein. Here, we only repeat that long-lived trace gases relevant for the stratosphere and mesosphere (such as CH4, N2O, and the CFCs) are prescribed as lower boundary condition [Kerkweg et al., 2006] on the basis of observations.
 The inclusion of NIR photolysis of HO2NO2 is essential to obtain agreement with the MIPAS observations, as can be seen from Figures 11 and 12. In this case the agreement between observations and simulations is within ±20% for most of the middle and lower stratosphere. This holds for zonal averages from 3 days in September 2002 (21–23 September 2002) (Figure 11) as well as snapshots' along a fixed latitude (63°S) across the split Antarctic vortex on 22 September 2002 (Figure 12). The most prominent effect of including the NIR photolysis in the model, leading to almost perfect agreement with MIPAS observations, is observed in the polar lower to middle stratosphere of both hemispheres, and in the (sub)tropical stratospheric HO2NO2 VMR peak.
 The model describes the overall zonal distribution very well, including the enhancement near the South pole. The cross section along 63°S reproduces the low values within the displaced polar vortex (in the Western Hemisphere between ≈135°W and 45°E) well, and also matches the higher VMRs outside the polar vortex in the East. Remaining differences (see Figure 11, middle right and middle bottom plots) are over most regions within the systematic uncertainty of the observational data (mainly determined by the uncertainty of spectroscopic data) and, to some extend, attributed to the lower vertical resolution of MIPAS observations. The lower vertical resolution of MIPAS HO2NO2 of 4 to 6 km results in a damping of the HO2NO2 VMR peak while its wings are broadened. This is exactly the pattern observed for the model-MIPAS differences (see Figure 11, middle right). The most pronounced differences are the higher HO2NO2 abundances above the stratospheric peak in MIPAS observations; these, however, are attributed to the rapidly deteriorating vertical resolution (see Figure 5) toward higher altitudes in the measurements which maps larger VMRs from the stratospheric peak to higher altitudes. This was confirmed by convolving the model output with MIPAS averaging kernels to adjust vertical resolution which reduced remaining differences between 10 hPa and 3 hPa (ca. 40 km; MIPAS HO2NO2 above this altitude level is considered to be driven by a priori information) to 2/3 of their unconvolved values. In particular, the difference maximum at ≈10 hPa and the threefold difference pattern disappeared. Further, above the HO2NO2 stratospheric maximum, the observational data suffer from the unresolved problems regarding the temperature dependence of the spectroscopic data (see section 3.2), and differences between model and observations in this altitude region should not been overinterpreted at the current state of data analysis. Remaining differences between model and observations could be due to uncertainties in rate constants, cross sections and quantum yields of reactions R1 to R4. This will be investigated in a future study. Control runs with deactivated NIR photolysis yield HO2NO2 abundances being up to a factor of two higher, with differences being dependent on latitude and altitude and most pronounced in the lower polar to midlatitude stratosphere of both hemispheres. In contrast, line list based retrievals (version 9) are even lower than the model results including NIR photolysis, in particular in the middle stratosphere where the HO2NO2 distribution has its maximum.
 With their model results, Evans et al.  provide an overview on the latitudinal distribution of HO2NO2 and its seasonal variation, both at the potential temperature level of 460 K (≈18 km), for the years 1997 to 1999. According to Evans et al. [2003, Figure 3], HO2NO2 is expected to have its maximum in Northern hemispheric spring in subpolar to middle latitudes, reaching VMRs of 240 pptv without including NIR photolysis, and around 100 pptv (reduction by 60%) when NIR photolysis is included. At 18 km, MIPAS observations provide HO2NO2 VMRs of 75 pptv for 40 to 60° N (see Figure 7b). Time-latitude cross sections at 460 K as presented by Evans et al. [2003, Figure 4] indicate highest HO2NO2 VMRs over the Arctic summer pole with values above 400 pptv (without including the NIR photolysis) and up to 190 pptv (when NIR photolysis is included). MIPAS HO2NO2 VMRs reach up to 130 pptv at 18 km, but up to 185 pptv 3 km higher (see Figures 13 and 7c). Northern polar winter values are predicted to be around 100 pptv without and 10 pptv with NIR photolysis considered, while MIPAS provides VMRs below 10 pptv in December 2002, and below 20 pptv in December 2003 (Figure 13). Over the Antarctic, the model without NIR photolysis predicts highest HO2NO2 VMRs during SH autumn with values up to 300 pptv; with NIR photolysis highest abundances are found during summer with values up to 125 pptv. MIPAS observations are highest during Southern Hemisphere summer (December/January) with abundances up to 140 pptv. In the tropics and over the winter poles, VMRs below 25 pptv are predicted by the model, similar to the values measured by MIPAS. In summary, as for the comparison with E5/M1, the MIPAS observations and the model results from Evans et al.  are in good agreement, if NIR photolysis is taken into account in the model, and if MIPAS uncertainties and the different years compared are considered.
 Comparisons between observations and model simulations were performed previously [Salawitch et al., 2002; Evans et al., 2003; Stowasser et al., 2002; Wetzel et al., 2002, 1997; Sen et al., 1998]. For all comparisons it is common that the model overestimated the observed HO2NO2 abundances if the NIR photolysis channel was not taken into account. These comparisons were improved by including the NIR photolysis path in the model calculations [Salawitch et al., 2002; Evans et al., 2003]. However, even then differences up to a factor of 2 remained. For example, the MkIV May 1997 profile which was compared to model simulations including NIR photolysis [Salawitch et al., 2002] was by a factor of 1.47 (at the altitude of the model profile's peak) lower than the model profile. The September 1993 profile in the same publication however, was consistent with the model result. Evans et al.  showed that the consistency between model profiles and observations was improved by inclusion of the NIR photolysis path into their model calculations. However, even then model profiles remained too high compared to MIPAS-B observations by factors of 1.2 to 2.6 (determined from Evans et al. [2003, Figure 2] at the “Run N1” model profile peaks). In each case except the September 1993 profile of MkIV, agreement between model and observations would be improved considerably by increasing the observational profiles by a factor of ≈1.5, i.e., if tabulated absorption cross sections would have been used instead of line list data. We take this as an additional indication that, by taking into account both the NIR photolysis path in the model calculations and the absorption cross section in spectral data analysis, discrepancies between models and observations can largely be resolved, and no major lack of understanding of the stratospheric HO2NO2 (photo)chemistry remains evident.
 We have presented the first global upper tropospheric and stratospheric data set of HO2NO2 from MIPAS/ENVISAT observations, covering the globe from pole to pole and a period of 19 months from September 2002 to March 2004, for daytime (near 1000 local time) and nighttime (near 2200 local time) conditions. Maximum stratospheric VMRs of 310 pptv are observed in northern summer subtropical to midlatitudes near 27 km, while lowest peak values occur in both hemispheres during polar night. Day-night differences are observed above 27 km only, providing a diurnal amplitude of a factor of 2 at these altitudes. MIPAS HO2NO2 data are available via http://www.fzk.de/imk/asf/ame/envisat-data for registered users.
 Significantly different results are obtained depending whether spectroscopic line list data [Brown et al., 1987; Rinsland et al., 1986] or absorption cross sections [May and Friedl, 1993] are used. Profiles retrieved on basis of spectroscopic absorption cross sections are ≈1.5 times higher. An open issue, still to be resolved, is the correct description of the temperature dependence of the HO2NO2 absorption cross sections. The tabulated absorption cross sections are available for 220 K only, while the line list provides constant lower-state energies, and appropriate information is missing to correctly compute the vibrational partition sums. This situation might improve if a detailed spectroscopic analysis, providing temperature and pressure dependence of the absorption cross sections of the HO2NO2ν6 Q branch, can be made available to the atmospheric spectroscopy community. Monthly zonal mean profiles of MIPAS/ENVISAT HO2NO2 have been compared to balloon profiles from the MkIV and MIPAS-B2 experiments measured prior to the MIPAS/ENVISAT mission, and no significant indication of inconsistencies has been found, provided the same spectroscopic data are used.
 HO2NO2 abundances measured with MIPAS and based on the tabulated absorption cross sections are in good general agreement with ECHAM5/MESSy1 model calculations with activated NIR photolysis path, thus confirming its importance. Control runs with deactivated NIR photolysis yield HO2NO2 abundances being up to a factor of two higher, with differences being dependent on latitude and altitude and most pronounced in the lower polar to midlatitude stratosphere of both hemispheres. The residual differences between model calculations including NIR photolysis and cross section-based observations are within the systematic errors of the measurements over wide latitude-altitude regions. Line list based retrievals are even lower than the model results including NIR photolysis, in particular in the middle stratosphere where the HO2NO2 distribution has its maximum. These results suggest that the major part of model-observation differences was caused by the analysis of the measurements rather than a lack of understanding of stratospheric chemistry. This conclusion does not only apply to ECHAM5/MESSy1 results, but also to other model calculations including the NIR photolysis as presented in the literature [e.g., Salawitch et al., 2002; Evans et al., 2003]. Nevertheless, there remains an urgent need for improved spectroscopic data, either as line data in an enhanced spectral range with correct description of the temperature dependence, or absorption cross-section data covering all relevant ranges of upper tropospheric/stratospheric pressures and temperatures.
 The HO2NO2 data set provided by MIPAS carries valuable information on the role of photolysis channels in various wavelength ranges, like that in NIR, as sinks for HO2NO2, and on the coupling of HO2NO2 to the HOx chemistry in the middle to lower stratosphere and upper troposphere. With its diurnal sampling, the MIPAS data set can also be used to constrain the diurnal variation of HO2NO2. Although MIPAS does not provide a full diurnal cycle, the annual and latitudinal variation of its observation time relative to sunrise and sunset times can be used to study the diurnal effects in the chemical processes.
 We thank ESA for provision of MIPAS level 1b data. The research work of members of the IMK-IAA MIPAS group has been funded by the EU via project TOPOZ-III (contract EVK2-CT-2001-00102) and BMBF via project 07 ATF 53 (SACADA). The IAA team has been partially supported by projects REN2001-3249/CLI and ESP2004-01556 and EC FEDER funds. We are grateful to Linda Brown and Randall Friedl, JPL, for helpful comments regarding the spectroscopic data. The authors would like to thank Geoffrey Toon and Bashwar Sen, JPL, for providing us with previously unpublished MkIV data and for their helpful and supportive comments. The constructive comments of Ross Salawitch and another anonymous reviewer are gratefully acknowledged.