Retrieval of stratospheric NOx from 5.3 and 6.2 μm nonlocal thermodynamic equilibrium emissions measured by Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on Envisat



[1] We present the first global observations of stratospheric NOx(= NO + NO2) from the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on Envisat during 24 July, 18 to 27 September, and 11 to 13 October 2002. Volume mixing ratio profiles of both NOx species were derived from MIPAS limb emission spectra by means of an innovative retrieval scheme under consideration of nonlocal thermodynamic equilibrium (non-LTE) effects. In the quiescent atmosphere, the estimated accuracy of retrieved NOx at the altitude of its stratospheric mixing ratio maximum at 35–40 km is around 1–2 ppbv, and the vertical resolution is around 3.5–6.5 km at altitudes between 20 and 50 km. In order to correctly consider NO2 non-LTE effects in the retrievals, the photochemical excitation rate of NO2(v3 > 0) vibrational states was derived from NO2(002→001) emissions and was found to be about 50 times smaller than previously estimated from Limb Infrared Monitor of the Stratosphere (LIMS) measurements. The NOx partitioning of the retrieved data is in excellent agreement with steady state photochemistry, which confirms predicted stratospheric NO(v > 0) non-LTE population enhancements. The retrieved NOx abundances are also consistent with Halogen Occultation Experiment (HALOE) NOx observations.

1. Introduction

[2] The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) [Fischer and Oelhaf, 1996; European Space Agency, 2000] was launched on board the Environmental Satellite (Envisat) into its Sun-synchronous polar orbit on 1 March 2002. MIPAS measures limb radiance spectra with high spectral resolution (0.035 cm−1) covering the midinfrared from 4.1 to 14.7 μm (i.e., 685–2410 cm−1), thus offering the opportunity to infer abundances of the active odd nitrogen species NO and NO2, from their NO(1 → 0) and NO23 → ν3 − 1) emissions at 5.3 and 6.2 μm, respectively.

[3] It has been reported that atmospheric NO emissions at 5.3 μm (i.e., 1800–1950 cm−1) depart from local thermodynamic equilibrium (LTE) down to stratospheric altitudes [Kaye and Kumer, 1987]. Kaye and Kumer pointed out that stratospheric daytime NO(v > 1) populations are enhanced compared to LTE because of vibrationally excited NO formed in photochemical reactions. However, the existence of stratospheric non-LTE excitations of NO has not yet been verified by means of atmospheric measurements.

[4] Thermospheric emissions of NO, which significantly contribute to the measured radiances at stratospheric tangent heights, show nonthermal behavior not only in the vibrational, but also in the rotational. and spin orbit structure [Armstrong et al., 1994; Lipson et al., 1994; Sharma et al., 1996; Funke and López-Puertas, 2000]. The need to consider these thermospheric non-LTE effects in the accurate retrieval of stratospheric nitric oxide from 5.3 μm emissions as measured by MIPAS has already been demonstrated [Funke et al., 2001].

[5] In the case of NO2, it is not clear to which degree the NO2(v3) levels, responsible for the emission near 6.2 μm (i.e., 1400–1650 cm−1), are enhanced by non-LTE processes in the daytime. On one hand, Kerridge and Remsberg [1989] found experimental evidence for non-LTE emissions at 6.2 μm from measurements taken by the Limb Infrared Monitor of the Stratosphere (LIMS) instrument on board the Nimbus 7 satellite. The chemiluminescence reaction

equation image

and the absorption of solar radiation at 400–800 nm are thought to be the major sources of non-LTE excitation of the NO2(v3 = 1–7) levels. Recently, Evans and Shepherd [1996] have reported the existence of a new airglow layer in the visible (500–1000 nm) located in the stratosphere. They propose that this emission originates from electronically excited NO2 created from the reaction of NO and O3. Radiative and collisional decay of this electronically excited NO2 is then thought to be the non-LTE excitation source of the infrared v3 levels in reaction (R1). The finding of Evans and Shepherd [1996] proves, at least, that the excitation source exists in the atmosphere, but whether it is strong enough to excite NO2(v3 = 1–7) or whether it is efficiently quenched before being transferred to or in the NO2(v3) levels is unclear. On the other hand, the analysis of the 6.2 μm emission measured by the Improved Stratospheric and Mesospheric Sounder (ISAMS) on the Upper Atmospheric Research Satellite has shown that these effects should be much smaller than those found from the analysis of LIMS measurements [Zaragoza et al., 1998]. In this sense, Remsberg et al. [2004] have recently reanalyzed LIMS radiances (V6) and have included better estimates of the orbital attitude for LIMS on Nimbus 7 which affects the NO2 retrieval, in particular its day/night difference. A reanalysis of the non-LTE hot bands of NO2 in LIMS has not been performed yet, but an initial inspection indicates that the day/night difference in NO2 (non-LTE enhancement) is no longer an issue with LIMS Version 6 (E. E. Remsberg, personal communication, 2004). Apart from the photochemical excitations, radiative processes may also generate non-LTE populations of the NO2(v3) levels under particular atmospheric conditions in the upper stratosphere and mesosphere. For example, in the polar winter mesosphere, NO23) excitation by absorption of the upwelling radiative flux is insufficient to compensate for spontaneous emission (due to the cold troposphere) which results in non-LTE populations with vibrational temperatures below the kinetic temperature [López-Puertas and Taylor, 2001]. The impact of non-LTE on the retrieval of stratospheric NO2 from MIPAS measurements depends strongly on the photochemical excitation rates of the NO2(v3) states which are still unknown. Because of its high spectral resolution, which allows for discrimination of different NO23) bands, MIPAS data offer an excellent opportunity to quantitatively analyze NO2 non-LTE excitations as required for the accurate retrieval of NO2 from limb emission spectra.

[6] An innovative non-LTE retrieval scheme [Funke et al., 2001] has been included in the scientific MIPAS data processor [von Clarmann et al., 2003b] developed at the Institut für Meteorologie und Klimaforschung (IMK) and at the Instituto de Astrofísica de Andalucía (IAA) which allows accurate inference of NOx(=NO + NO2) volume mixing ratios (VMRs) under consideration of all important non-LTE processes mentioned before. It further enables the retrieval of NO2 non-LTE excitation rates as a preparative step before the retrieval of NO2 abundances.

[7] NOx plays an important role in controlling the stratospheric O3 abundance, directly by destruction of O3 within the catalytic NOx cycle

equation image
equation image

and indirectly by reducing the efficiency of O3 destruction in the ClOx cycle because of the chemical recombination of ClO and NO2 which gives ClONO2. Heterogeneous reactions involving N2O5 on polar stratospheric clouds (PSCs) may shift NOx to the reservoir species HNO3, making less NO2 available to react with ClO. This deactivation of NOx by heterogeneous chemistry is implicated in the large ozone losses over the polar regions in winter or spring. Non-LTE retrieval of NOx complements the measurements of other odd nitrogen species derived from MIPAS data (i.e., HNO3, HNO4, ClONO2, and N2O5) [Stiller et al., 2004; Höpfner et al., 2004; Mengistu Tsidu et al., 2004], enabling for the first time a complete determination of the NOy partitioning (except for NO3 and BrONO2) in the stratosphere with global coverage.

[8] Apart from the reaction of N2O and O(1D) in the stratosphere, odd nitrogen is formed in the thermosphere by dissociation of molecular nitrogen due to solar radiation and energetic particles. During polar winter, thermospheric NOx can be transported to the stratosphere [Siskind et al., 1997] where it serves as an additional source of stratospheric NOy. Up to now, the quantitative analysis of the impact of this additional source to the global NOy budget was limited because of the lack of NOx measurements during polar night [Siskind, 2000]. The pole-to-pole coverage of MIPAS NOx measurements, which are independent on illumination, will make an important contribution to quantify the thermospheric NOy sources.

[9] Stratospheric NOx has been measured since 1970 by ground-, balloon-, and aircraft-based measurements using in situ, radiometer and spectrometer techniques [i.e., Noxon et al., 1979; Kondo et al., 1985; Pfeilsticker et al., 1999]. Spaceborne solar occultation observations of NO and NO2 were performed in April 1985 by the Atmospheric Trace Molecule Spectroscopy instrument (ATMOS) on board Spacelab 3 [Russell et al., 1988] and later, on board the space shuttle, in March 1992, April 1993 and November 1994 [Newchurch et al., 1996]. Since 1991, the Halogen Occultation Experiment (HALOE) on the Upper Atmospheric Research Satellite (UARS) [Russell et al., 1993] is measuring continuously NO and NO2. Stratospheric NO2 radiometer observations of atmospheric limb radiances were made from 1978 to 1979 by the LIMS instrument [Gille and Russell, 1984], however, without taking into account possible NO2 non-LTE effects. The Improved Stratospheric and Mesospheric Sounder (ISAMS) on UARS [Taylor et al., 1993] measured NO 5.3 μm and NO2 6.2 μm limb radiances by means of radiometer technique. NO2 occultation measurements in the visible were performed by the Stratospheric Aerosol and Gas Experiment (SAGE II) [Cunnold et al., 1991] since 1984 and by the Polar Ozone and Aerosol Measurement (POAM II) Instrument [Randall et al., 1998] from 1993 to 1996. The follow-up instruments POAM III and SAGE III launched in 1998 and 2001, respectively, are operational in time of this writing. Stratospheric NO2 observations were also made by the Improved Limb Atmospheric Spectrometer (ILAS) and ILAS-II using solar occultation technique [Irie et al., 2002]. Simultaneous measurements of NO and NO2 from MIPAS on Envisat as described in this paper are the first of its kind using spectrally resolved limb emissions which allow to infer global NOx abundances independent on solar illumination.

[10] In section 2 we describe the IMK/IAA non-LTE data processor and characterize the retrieval of NO2 non-LTE excitation rates and of NOx VMR profiles from MIPAS/Envisat data in sections 36. The consistency of our NOx retrievals with theory and correlative measurements is assessed in section 7.

2. Experiment and Retrieval Method

[11] MIPAS on Envisat is a limb-viewing Fourier transform infrared spectrometer with 0.035 cm−1 spectral resolution, measuring atmospheric radiance spectra in the frequency range from 685 to 2410 cm−1 in five spectral channels [Fischer and Oelhaf, 1996; European Space Agency, 2000]. The field of view (FOV) is 30 km in the horizontal and approximately 3 km in the vertical direction. The nominal observation mode scans the limb in 17 sweeps, covering tangent altitudes from 8 to 68 km in 3 km steps up to 47 km, followed by sweeps at 52, 60, and 68 km. Flown on a Sun-synchronous orbit of 98.55 inclination at approximately 800 km altitude, MIPAS passes the equator in south direction at 1000 LT 14.3 times a day. During each orbit approximately 72 limb scans are recorded. MIPAS data of 24 July, 18–27 September, and 11–13 October have been used in this work, including about 6000 elevation scans. From 29 September until 10 October 2002 the instrument was switched off because of a failure of the stirling cooler. Extensive CPU time required for the non-LTE data analysis did not allow for further extension of the data set presented here. The L1B processing of the data (version 4.55), including all processing steps from raw data to calibrated spectra, has been performed by the European Space Agency (ESA) [Nett et al., 2002].

[12] The first step in the L2 processing of the L1B spectra was the determination of the spectral shift, followed by the retrieval of temperature and elevation pointing [von Clarmann et al., 2003a]. In the latter step, pressure is implicitly determined by means of hydrostatic equilibrium. Prior to the retrievals performed in this work, VMR profiles of H2O, O3, and ClO were obtained from MIPAS spectra.

[13] The non-LTE retrieval of NOx VMRs and NO2 photochemical excitation rates are performed with the IMK-IAA data processor [von Clarmann et al., 2003b], which is based on a constrained nonlinear least squares algorithm with Levenberg-Marquardt damping. Modeled spectra are iteratively fitted to the measurements by updating the actual vector of retrieval parameters xi of iteration i by

equation image

where ymeas is the vector of measurements and Sy is the measurement noise covariance matrix. K is the Jacobian matrix containing the partial derivatives ∂y(xi)/∂xi, KT its transposed, and R is a regularization matrix. λI (scalar times unity) is a the Levenberg-Marquardt damping term [Levenberg, 1944; Marquardt, 1963] which is forced to be zero in the last iteration. y(xi) is the result of the nonlinear radiative transfer model and xa is a vector containing the a priori knowledge of the unknown parameters, also used as initial guess. In all applications reported here, the vector of retrieval parameters includes, in addition to the target parameters, a continuum-like optical depth profile and a height-independent radiance offset as described by von Clarmann et al. [2003a]. NO and NO2 abundances as well as the NO2 photochemical excitation rates are retrieved on a height grid of 1 km altitude spacing. In order to stabilize the retrieval, a first-order Tikhonov [1963]-type regularization is applied [Steck, 2002]. The altitude-dependent regularization strength is chosen as a trade-off between vertical resolution and precision of the retrieved parameters.

[14] The generic non-LTE population model GRANADA [Funke et al., 2002] is integrated in the retrieval scheme in order to calculate the non-LTE populations of atmospheric emitters for the updated vector of retrieval parameters within each iteration step. This model calculates non-LTE populations of vibrational, rotational, and spin orbit states of all relevant atmospheric gases emitting in the infrared. Non-LTE radiative transfer modeling of the spectra and the Jacobian matrix is performed by the Karlsruhe Optimized and Precise Radiative Transfer Algorithm (KOPRA) [Stiller et al., 2002], using the non-LTE populations calculated by GRANADA. The dependence of non-LTE populations n on the retrieval parameters is considered by correcting the Jacobian matrix such that

equation image

Here, ∂y/∂n are the partial derivatives with respect to the non-LTE populations calculated by KOPRA, while (∂n/∂x) is provided by the non-LTE population model. This non-LTE correction of the Jacobian matrix is indispensable for the retrieval of specific non-LTE parameters such as the NO2 photochemical excitation rates, where the first term in equation (2) is zero.

[15] For reasons of efficiency and in order to reduce systematic errors, a microwindow approach has been chosen; that is, the retrieval is performed on small subsets of the measured spectra. Microwindows are selected during a preprocessing step using an optimizing selection tool [Echle et al., 2000].

[16] Characterization and diagnostics of the retrieval results are performed in linear approximation, i.e., the mapping of a variation of the measurement y into the vector of retrieved parameters x is described by the Gain matrix [Rodgers, 2000]

equation image

Within this formulation, retrieval errors due to instrumental noise are described by the random error covariance matrix

equation image

Retrieval errors due to uncertainties ε of the priori parameters ai used in the radiative transfer and non-LTE modeling are estimated by

equation image

The effect of regularization on the retrieval result is given by the averaging kernel matrix

equation image

Its elements, Aij, describe the response of the retrieved parameter j to a change of the “true” parameter i. The vertical resolution of the retrieved profile can thus be expressed by the full width at half maximum of the columns of A.

3. Retrieval of NO2(v3) Excitation Rates

[17] The determination of NO2(v3) non-LTE excitation rates is crucial for the retrieval of atmospheric NO2 abundances from emission measurements near 6.2 μm. As discussed in the introduction, these rates are poorly known. Simulations of stratospheric NO2 emissions in MIPAS spectra at 6.2 μm using minimum and maximum estimates for these rates range from ∼3 nW/(cm2 sr cm−1), a value well below the detection limit of MIPAS, up to 30 nW/(cm2 sr cm−1) [López-Puertas, 1997]. A detailed discussion on the possible NO2 excitation mechanisms is given by López-Puertas and Taylor [2001]. In summary, the two most important processes potentially exciting NO2 are the chemiluminescence reaction of NO and O3 (reaction (R1)) and fluorescence following absorption of solar radiation in the visible (400–800 nm). These two processes release sufficient energy to produce excited NO2 up to v3 = 7. As described by López-Puertas and Taylor [2001], the NO23) excitation rate ℰ due to both mechanisms consistent with the analysis of the LIMS data [Kerridge and Remsberg, 1989] can then be approximated at stratospheric altitudes by

equation image

where Jexc is the photoexcitation coefficient of NO2 and k2 is the rate coefficient of reaction (R2).

[18] In this work we retrieve the photochemical excitation rates ℰexp for NO2(v3) levels from the measured non-LTE emissions of the 002 → 001 hot band around 1600 cm−1. In order to constrain the NO2 density required for the non-LTE modeling within the retrieval, the NO2 abundance was retrieved simultaneously from the NO23) fundamental band.

[19] Within the retrieval, the non-LTE populations of the NO2(v3 = 1–7) levels are calculated by the GRANADA non-LTE model. Except for photochemical excitation, it includes all processes and rate coefficients as described by López-Puertas and Taylor [2001]. It is assumed that all photochemical excitation corresponding to the retrieved rates ℰexp goes initially to the NO2(v3 = 7) state. The de-excitation processes include radiative emission and collisional quenching with N2 and O2. The values for the collisional quenching rates are very similar to those used in the LIMS analysis.

[20] Microwindows used for the analysis cover the spectral range from 1570 to 1607 cm−1. Spectroscopic line data for the NO23) bands have been taken from Flaud et al. [2003]. Following the method of Steck [2002], the altitude-dependent strength of the regularization was chosen such that the precision of ℰexp is 2 × 106 cm−3s−1 which is around 10% of ℰmodel for midlatitude daytime conditions. The retrieval of ℰexp is sensitive between 25 and 35 km altitudes, however, without gaining substantial information on its profile shape.

[21] Around 500 scan profiles with global coverage have been analyzed in order to infer ℰexp which are most sensitive to photochemical excitation. As single scan results are generally below the noise-induced retrieval error of approximately 2 × 106 cm−3 s−1, a statistical data analysis is required. The mean value for ℰexp of all daytime measurements at 30 km is 0.42 × 106 cm−3s−1 while the nighttime mean value is close to zero. This is in accordance with the assumption that the chemiluminescence reaction of NO and O3 (reaction (R1)) and NO2 fluorescence are the main excitation pathways, since both processes are negligible at nighttime. Model calculations performed with the parameterization of both excitation mechanisms as defined in equation (7), using the NO2, O3, and NO abundances retrieved from MIPAS, show a daytime mean value for ℰmodel of 2.1 × 107 cm−3 s−1, which is around 50 times larger than ℰexp retrieved here. The standard deviation of both, day and nightttime measurements is around 1.7 × 106 cm−3 s−1 which is in accordance with the estimated noise-induced retrieval error for a single scan.

[22] The correlation between ℰexp and ℰmodel is demonstrated by Figure 1, where mean values for all measurements, whose corresponding model rates are within bins of 10 × 106 cm−3s−1, are plotted versus the mean values of the model rate bins. The errors of the mean values for ℰexp are estimated from the noise-induced retrieval errors for a single scan divided by equation image, where n is the number of measurements within the bin. In a similar way, errors of the model rate mean values are estimated as the quadratic sum of the retrieval errors for NO, NO2, and O3 divided by equation image. A regression fit with a slope of 0.02 indicates that the NO2(v3) excitation rate can be expressed by the parameterization given by equation (7) but with a reduction factor of 0.020 ± 0.004.

Figure 1.

Measured NO2(v3) photochemical excitation rates versus modeled rates at 30 km. Model calculations include excitation due to chemiluminescence (reaction (R1)) and fluorescence as described by López-Puertas and Taylor [2001] using rate coefficients in accordance with the LIMS analysis and the retrieved profiles of NO2, O3, and NO. Results of individual scans are indicated by solid circles (night) and squares (day). Mean values of measurements, whose corresponding model rates are within bins of 10 × 106 cm−3 s−1, versus the mean values of the model rates within the bins are shown with error bars. Vertical error bars refer to retrieval errors of the measurement mean values, horizontal error bars represent uncertainties of the model mean values due to retrieval errors of NO2, O3, and NO as defined in section 3. A regression fit with a slope of 0.02 is shown by a dotted line.

[23] A discrimination between both excitation path ways, chemiluminescence and fluorescence from MIPAS data is extremely difficult, since both terms in equation (7) are linearly dependent in photochemical equilibrium.

[24] The populations of the NO2(v3 ≥ 2) levels are mainly controlled by the photochemical excitation rate and by the collisional quenching with N2 and O2. In the retrieval of ℰexp we included values for the collisional quenching rates very similar to those used in the LIMS analysis. However, these quenching rates are not very well known and their uncertainties would directly map into the retrieved photoexcitation rates. A detailed review of the vibrational relaxation rates of NO2(v3) by N2 and O2 is given by López-Puertas [1997]. In summary, they have been measured by a number of techniques and theoretical studies including those of Golde and Kaufman [1974], Hui and Cool [1978], McAndrew et al. [1989], Adler-Golden [1989], Zuev and Starikovskii [1991], and Mazely et al. [1994]. The scatter of the rates measured for high (v3 ≥ 2) levels is rather high and can reach values a factor of 2–3 larger (depending on the state) than those used here (and used in the LIMS analysis). The use of the larger quenching rates in the retrieval would lead to increased values of ℰexp, approximately by same factor. It is important to note, however, that this would have a minor effect on the non-LTE populations of the NO2(v3) levels (the more efficient quenching compensates for the faster excitation rate) and hence would have very little impact on the retrieval of NO2 abundances.

[25] In order to prove whether the low excitation rate retrieved from the 002→001 band is also consistent with measured emissions from the high energetic NO2(v3 > 4) levels, an analysis of the 1450–1500 cm−1 region of 250 coadded MIPAS spectra has been performed. No NO2(v3 > 4) hot band emissions could be identified within the measurement noise of 0.7 nW/(cm2 sr cm−1) of the coadded spectra. These emissions would already exceed the detection limit when assuming a NO23) chemical excitation rate 13 times smaller than ℰmodel.

4. Retrieval of NO2

[26] The NO2 VMR is retrieved from the 001→000 band emissions using 14 microwindows between 1590 and 1635 cm−1. The non-LTE modeling of these emissions within the retrieval includes the retrieved photochemical excitation rate of the NO2(v3) states and all de-excitation processes as described in section 3. Radiative excitation of the NO2(001) state by absorption of upwelling tropospheric radiation is modeled by GRANADA taking into account emissions of water vapor, methane, molecular oxygen continuum, and NO2 itself. Attenuation of the upwelling flux due to clouds is included whenever clouds have been detected in the MIPAS line of sight using the detection method proposed by Spang et al. [2001]. Hereby we assume that the cloud top height is equal to the highest cloud-contaminated tangent height.

[27] A characterization of the NO2 retrievals has been performed on basis of a set of 15 representative single scans covering different atmospheric conditions (i.e., tropics, midlatitudes, polar regions) and illumination. For illustration, Figure 2 shows a typical nighttime VMR profile of NO2 retrieved from MIPAS measurements at tropical latitudes (10.9°N) taken on 20 September 2002, the corresponding averaging kernel and the vertical resolution. In all cases investigated, the averaging kernels indicate that the measurements contain significant information on the NO2 abundances between 15 and 65 km. Generally, around 9–11 degrees of freedom, i.e., points of independent altitude information, are calculated from the trace of the averaging kernel matrix. The vertical resolution of the retrieved NO2 profiles is around 3.5–6.5 km in the 16–50 km altitude range and 9–12 km below and above. No significant difference between daytime and nighttime vertical resolutions was found.

Figure 2.

(top) Retrieved nighttime NO2 VMR profile (solid line) along with first guess profile (dotted) and random errors bars from MIPAS measurements taken on 20 September 2002, 1902:52 UT at 10.9°N 47.7°E. (bottom) Columns of the (left) corresponding averaging kernel and (right) vertical resolution.

[28] A complete error budget for the NO2 retrieval including random errors due to instrumental noise and all important systematic error sources is given in Table 1. The total NO2 retrieval errors are generally below 1 ppbv, except for altitudes around 35 to 40 km, where systematic errors, particularly due to uncertainties of the spectroscopic data used, are dominating. Since these errors scale with the amount of NO2, higher retrieval errors can be found at nighttime compared to daytime due to diurnal variations of NO2. The original analysis of the NO2 data set suffered from a latitude-independent negative bias of 5–10% above 30 km and a positive bias below (up to 40% at 20 km) due to inappropriate instrumental line shape (ILS) modeling. Profiles presented here, however, rely on data processing based on a revised ILS model (F. Hase et al., personal communication, 2003). Random retrieval errors estimated from instrumental noise contribute in all altitudes with 0.2–0.3 ppbv. Systematic errors in the region above about 50 km are also introduced by the uncertainties in the collisional rates used in the non-LTE modeling. The most uncertain rate is the collisional quenching of NO2(001) by N2 and O2. This rate has not been measured in the laboratory for the atmospheric temperatures. The value used here was derived from the measurements of the quenching rates for higher v3 states (see section 3) and the theoretical energy dependency proposed by Adler-Golden [1989]. From the scatter of the measurements of high v3 levels and the upper limit for the deactivation of NO2(010) measured by Toselli et al. [1990] we estimate an uncertainty of 100% in quenching rate of NO2(001). This leads to a maximum error in the retrieved NO2 VMR of about 10%. In summary, the accuracy of retrieved NO2 is in the order of 7–10% at 40 km VMR peak altitude during night and 15–20% at 30 km VMR peak altitude during day.

Table 1. Error Budget at Selected Altitudes for the Retrieval of Stratospheric Daytime NO2a
Height, kmTotal ErrorbInstrumental NoiseInterfering GasescTemperaturedPointingeSpectroscopic DatafGaingSpectral ShifthNon-LTEi
  • a

    Errors are given in absolute units, ppbv. Values in parentheses correspond to nighttime whenever different to daytime.

  • b

    Defined as quadratic sum of all individual errors.

  • c

    Variability of the interfering gases not retrieved from the same data set is assumed on basis of their climatological variability.

  • d

    Based on temperature uncertainty of 2 K.

  • e

    Based on tangent altitude uncertainty of 150 m.

  • f

    Based on uncertainty of spectroscopic data of 10% (worst case) for NO2 information by J. M. Flaud (personal communication, 2003).

  • g

    Based on gain calibration error of 1%.

  • h

    Based on a residual spectral shift error of 0.0005 cm−1.

  • i

    Uncertainties of parameter driving non-LTE processes (photochemical excitation with 15% as determined in this study and collisional quenching with 100%).<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.1<0.10.1<0.1<0.1<0.1 (0.3)0.2 (0.1)<0.1<0.1<0.1
35.00.9 (1.5) (0.5)0.1 (0.3)0.7 (1.3)0.2<0.1<0.1
40.00.6 (1.2)0.3<0.10.3 (0.5)0.1 (0.2)0.4 (1.0)0.1<0.1<0.1
45.00.3 (0.5)0.2<0.1<0.10.1 (0.2)0.2 (0.4)<0.1<0.1<0.1 (0.2)
50.00.3 (0.6)0.2<0.1<<0.1<0.1<0.1 (0.5)
55.00.3 (0.6)0.2<0.1<<0.1<0.1<0.1 (0.5)
60.00.3 (0.4)0.2<0.1<<0.1<0.1<0.1 (0.2)
65.00.3 (0.4)0.2<0.1<<0.1<0.1<0.1 (0.2)

[29] As an example, the distribution of NO2 along MIPAS orbit 2915 recorded on 20 September 2002 is shown in Figure 3. The retrieved abundances show all the features of a typical NO2 global distribution such as a smaller NO2 concentrations during day (profiles 12–46) due to photochemical conversion into NO, and the generally higher NO2 abundances in the tropics compared to polar latitudes.

Figure 3.

(top) Retrieved (left) NO2 and (right) NO VMRs along MIPAS orbit 2915 measured on 20 September 2002, versus altitude and profile number/latitude. The daytime part of this orbit covers profiles 12–46, crossing the equator at 120°E longitude. (bottom) Non-LTE induced enhancement of the vibrational populations of (left) NO2(001) and (right) NO(1), whose emissions are used for the retrieval of NO2 and NO, respectively. White areas represent values beyond the scale range. Note that some regions of non-LTE enhancements higher than the scale range are caused by VMRs close to zero (i.e., NO2 around 50 km and NO below 30 km during the day).

[30] As a consequence of the low photochemical excitation rates of NO2(v3) previously determined, the populations of the measured NO2(001) state are generally close to LTE within 2–3% below 50 km (see Figure 3, bottom left). Above 50 km, however, NO2(001) populations can be up to 30% lower than in LTE due to insufficient collisional and radiative excitation. This effect is most pronounced at polar winter, where temperatures in the upper stratosphere/lower mesosphere are rather high. The omission of non-LTE within the NO2 retrieval would thus lead to an underestimation of upper stratospheric NO2 up to 30% for these conditions.

5. Retrieval of NO

[31] Once NO2 is derived, NO VMRs are retrieved from its 1→0 fundamental band emission using 14 microwindows in the 1845–1915 cm−1 region. The non-LTE modeling of these 5.3 μm emissions is performed as described by Funke and López-Puertas [2000] considering vibrational states up to v = 3 and rotational states of both spin orbit components up to J = 35.5. In the Sun-lit stratosphere, chemical production of excited NO(v > 0) states by

equation image

and by the reaction of NO2 with atomic oxygen (reaction (R3)) may lead to a considerable enhancement of 5.3 μm emissions with respect to LTE. The corresponding chemical excitation rates are calculated using the NO2 abundances retrieved previously from the same data set. Atomic oxygen abundances below 70 km are derived from retrieved O3 assuming photochemical equilibrium under consideration of the first 4 reactions of Table 2. Photolysis rates of O3 and NO2 needed hereby are calculated by means of the Tropospheric Ultraviolet-Visible (TUV) radiation model [Madronich and Flocke, 1998] under consideration of the atmospheric conditions and illumination of each measurement. Radiative excitation of NO due to absorption of upwelling radiation is considered in the GRANADA non-LTE model in a similar way as for NO2.

Table 2. Chemical Reactions and Rates Included in the Photochemical Equilibrium Model for Atomic Oxygen and NOx
ReactionRates,a cm3 s−1Reference
  • a

    For three-body reactions units are cm6 s−1.

O3 + hν → O2 + OTUV version 4.2Madronich and Flocke [1998]
O3 + H → O2 + OH1.4 × 10−10 exp(−470/T)Sander et al. [2003]
O3 + O → 2O28.0 × 10−12 exp(−2060/T)Sander et al. [2003]
O + O2 + M → O3 + M6.4 × 10−34 exp(T/300)−2.4Sander et al. [2003]
NO2 + hν → NO + OTUV version 4.2Madronich and Flocke [1998]
NO2 + O → NO +O25.6 × 10−12 exp(180/T)Sander et al. [2003]
NO + O + M → NO2 + M9.0 × 10−31 exp(T/300)−1.5Sander et al. [2003]
NO + O3 → NO2 + O23.0 × 10−12 exp(−1500/T)Sander et al. [2003]
NO + ClO → NO2 + Cl6.4 × 10−12 exp(290/T)Sander et al. [2003]
NO + HO2 → NO2 + OH3.5 × 10−12 exp(250/T)Sander et al. [2003]

[32] Since thermospheric non-LTE emissions of NO contribute with up to 50% to the spectra recorded at stratospheric tangent heights, the model atmosphere, usually covering altitudes from the ground to 120 km, had to be extended up to 200 km. The NRLMSISE-00 empirical model of the atmosphere [Picone et al., 2002] is used from 70 km to 200 km. Vibrational, rotational, and spin orbit non-LTE in the thermosphere are mainly driven by quenching with atomic oxygen and chemical excitation of NO due to

equation image

The modeling of these processes requires concentrations of thermospheric O, O2, and N, which are all taken from the NRLMSIS-00 model.

[33] Although MIPAS scans the limb only up to tangent heights of 70 km in the nominal observation mode, the NO retrieval profile grid covers altitudes up to 200 km in order to properly account for the highly variable amount of thermospheric NO in the MIPAS line of sight. It is evident that no substantial information on the thermospheric NO profile shape can be extracted from nominal MIPAS observations. The Tikhonov-type regularization [Tikhonov, 1963] in an implementation as suggested by Steck [2002] hence constrains the retrieved profile in the thermosphere to the a priori profile shape. However, errors in the a priori profile shape are not significantly propagated down to the stratosphere. Spectral measurements with tangent altitudes below 20 km have been ignored for the NO retrievals in order to avoid upward error propagation.

[34] A characterization of the NO retrievals has been performed in the same manner as for NO2 on basis of a set of representative scans covering different atmospheric conditions. As an example, Figure 4 shows a typical stratospheric daytime NO VMR profile retrieved from MIPAS measurements at 36.4°N taken on 20 September 2002. In all cases, averaging kernels indicate that the measurements contain significant information on the NO abundances between 22 and 60 km. Between 20 and 50 km, the vertical resolution is around 4–7 km. Table 3 summarizes the main error contributions for the stratospheric NO retrieval. In general, the precision of the retrieved NO estimated from instrumental noise is better than 1 ppbv. The most important error source for the retrieval of stratospheric NO is the horizontal variability of thermospheric NO due to the fact that the instrument line of sight passes the thermosphere at different points with up to 3000 km distance. An assessment of retrieval errors caused by thermospheric gradients has been performed by analyzing the nighttime retrievals. As the NO concentration is negligibly small up to around 60 km during nighttime, any deviation of the retrieved profile should be caused either by systematic or noise-induced retrieval errors. Assuming that thermospheric gradients are the dominant error source at night (other important error sources as uncertainties of spectroscopic data scale with the amount of NO), a statistical evaluation of nighttime retrievals allows for the quantification of the resulting retrieval errors. These were found to be up to 5 ppbv around 50 km at polar latitudes (around 70° geomagnetic latitude) where latitudinal gradients of thermospheric NO are most pronounced because of auroral production. At lower latitudes, however, NO retrieval error due to thermospheric gradients are generally less than 1 ppbv. Other important error sources are uncertainties in temperature and spectroscopic data, which contribute up to 0.4 and 0.7 ppbv, respectively, at 40 km during the day. A latitude-independent negative bias of 5–10% between 35 and 45 km and a high bias of around 20–40% in the remaining altitude regions due to inappropriate ILS modeling have been identified in the original NO data. However, profiles presented in this work have been reprocessed using a revised ILS model (F. Hase et al., personal communication, 2003). In summary, at 40 km VMR peak altitude, the accuracy of retrieved daytime NO is in the order of 8–15% at quiescent conditions but degrades to 20–40% during auroral activity.

Figure 4.

(top) Retrieved daytime NO VMR profile (solid line) along with first guess profile (dotted) and random error bars from MIPAS measurements taken on 20 September 2002, 1759:34 UT at 36.4°N 115.9°W. (bottom) Columns of the (left) corresponding averaging kernel and (right) vertical resolution.

Table 3. Error Budget at Selected Altitudes for the Retrieval of Stratospheric Daytime NOa
Height kmTotal ErrorbInstrument NoiseInterface GasescTemperaturedPointingeSpectroscopic DatafGaingThermospheric GradientshNon-LTEi
  • a

    Errors in absolute units, ppbv. Errors for polar conditions are given in parentheses whenever different to midlatitude conditions.

  • b

    Defined as quadratic sum of all individual errors.

  • c

    Variability of the interfering gases not retrieved from the same data set, which is assumed on basis of their climatological variability.

  • d

    Based on temperature uncertainty of 2 K below 80 km and 50 K above.

  • e

    Based on tangent altitude uncertainty of 150 m.

  • f

    Based on uncertainty of spectroscopic data of 10% (worst case) for NO information by J. M. Flaud (personal communication, 2003).

  • g

    Based on gain calibration error of 1%.

  • h

    Estimated errors due to thermospheric horizontal inhomogeneities.

  • i

    Uncertainties of parameter driving non-LTE processes based on Funke et al. [2001].<0.10.1<0.10.1<0.1<0.1<0.1
25.00.6 (0.7)0.5<<0.10.1 (0.3)0.2
30.01.1 (1.4)1.0<<0.10.1 (0.8)0.2
35.01.3 (2.0)1.0< (1.5)0.3
40.01.4 (2.3)1.0< (1.8)0.4
45.01.8 (3.2)1.0< (2.8)0.5
50.01.5 (4.7)1.0<<0.10.7 (4.5)0.4
55.01.3 (3.2)1.0<<0.10.5 (3.0)0.2
60.00.7 (1.2)0.5<<0.10.2 (1.0)0.2

[35] Figure 3 (top right) shows the distribution of NO along MIPAS orbit 2915. As expected, retrieved NO abundances have values within the noise during the night. Daytime NO show high values in the tropics and over the north pole while less NO is measured in high southern latitudes.

[36] Chemical excitation of NO due to reactions (R3) and (R4) causes a NO(v = 1) population enhancement of around 40–60% in the 25–40 km region during daytime (see Figure 3, bottom right), resulting in up to 6 ppbv smaller NO VMRs retrieved under consideration of non-LTE compared to LTE.

6. Photochemical Constraint for Retrieval at Twilight Conditions

[37] Because of the large depth of air masses observed along the line of sight (400–500 km within the tangent layer), the retrieval of NO and NO2 at twilight conditions have to deal with pronounced horizontal gradients caused by the fast photochemistry which converts NO into NO2 and vice versa. The photochemically induced horizontal inhomogeneities along the instrument's line of sight have been taken into account for scans close to the terminator (solar zenith angles of 80–100°) by means of a simple photochemical equilibrium model which constrains the NOx partitioning Q(s) = [NO(s)]/[NO2 (s)] along the optical path s of the instrument-dependent on the solar zenith angle at s. Assuming a homogeneous horizontal distribution of NOx, the NO and NO2 abundances at the optical path segment s passing a given altitude level are scaled with respect to the actual abundance at the center of the scan s0 and the same altitude level as

equation image
equation image

within the forward calculations of the retrieval. The photochemical equilibrium model includes the chemical reactions as summarized in Table 2. This requires photoabsorption coefficients for NO2 and O3 provided by the TUV model (version 4.2), pressure, temperature, and abundances of O3, ClO, H, and HO2 as input. Pressure, temperature, O3 and ClO abundances are used as retrieved from the same MIPAS data [von Clarmann et al., 2003a; Glatthor et al., 2004, 2005], while concentrations of H and HO2 are taken from a chemical transport model calculation [Garcia, 1983].

7. Consistency Tests

[38] Since this is the first time that non-LTE has been included in the retrieval of stratospheric NO and NO2 from atmospheric emission measurements, a verification of the retrieval method and the non-LTE modeling included was necessary. Here, this has been addressed by intercomparison of the retrieval results to model calculations and correlative measurements. These intercomparisons were not meant to substitute a detailed validation of the data products which is beyond the scope of this work and will be addressed in a future paper.

[39] During the day, NO and NO2 are in photochemical equilibrium in the stratosphere up to approximately 50 km. The same photochemical model as used for constraining the horizontal distribution of NO and NO2 along the line of sight at twilight conditions (see section 6) can then be used to prove the consistency of retrieved NO and NO2 with theory. Figure 5 shows a comparison of retrieved daytime NO from MIPAS orbit 2915 measured on 20 September 2002 (see also Figure 3) to model calculations using NO2, as well as pressure, temperature, O3, and ClO retrieved from the same data set. Below 35 km, the mean value of all NO measurements coincides with the model calculations within 0.5 ppbv. Above, differences reach up to 1.5 ppbv, however still within the 1σ standard deviation of single scan differences. The omission of stratospheric non-LTE excitations of NO due to reactions (R3) and (R4) in the retrieval results in an overestimation of derived NO of up to 6 ppbv at 35 km compared to the model calculations. The excellent agreement of measured and steady state NOx partitioning clearly confirms the predicted stratospheric non-LTE excitations of NO and it further reflects the self-consistency of retrieved NOx data products.

Figure 5.

Comparison of measured NO (all daytime retrieval from MIPAS orbit 2915) to model calculations assuming photochemical equilibrium. Calculations have been performed using the retrieved pressure, temperature, O3, NO2, and ClO from MIPAS (see text for further details). (left) Mean values and 1σ variances (bars) of the differences of retrieved and modeled NO VMRs versus altitude. NO retrieval including the chemical excitation as non-LTE source in the stratosphere are shown by the solid curve, NO retrieval without considering this non-LTE process are shown by the dashed curve. (right) Scatterplot of the individual measurements versus model calculations for all altitudes including (solid circles) and not including (open circles) chemical excitation.

[40] HALOE solar occultation measurements of NO and NO2 offer the opportunity to compare our NOx data products with data not affected by non-LTE and thus allow for a further verification of the non-LTE retrieval approach used here. However, the rapid photochemical conversion of both species and the nature of HALOE measurements (solar occultation) make the comparison difficult. Thus profiles of total NOx VMR rather than NO and NO2 profiles have been compared. In addition, diurnal variations of NOx caused by conversion to N2O5 at nighttime and recovery by N2O5 photolysis during the day have to be taken into account when comparing NOx measurements of both instruments taken at different solar local times. During night, the buildup of N2O5 from NO2 takes place by the reactions

equation image
equation image

where reaction (R6) is limiting the rate for the N2O5 formation. As in the dark the only known loss process of N2O5 is thermal decomposition, which in the stratosphere is very slow, the time evolution of the NOx decrease during night can be characterized by

equation image

Here, t0 is the time at sunset or sunrise, and k6 is the rate constant for reaction (R6) taken from Sander et al. [2003]. For the period of 22 to 29 September 2002, 52 coincidences between MIPAS (nighttime) and HALOE V19 (sunset and sunrise) measurements have been found within a latitude × longitude box of 3° × 15° and a maximum time lag of 12 hours. HALOE NOx measurements have been corrected for diurnal variations according to equation (10) using colocated HALOE V19 O3. The averaged differences between MIPAS and HALOE NOx are smaller than 20% between 25 and 50 km (see Figure 6), however, biased above 30 km toward higher MIPAS NOx. Given that HALOE NO measurements show a low bias of 15–30% above 25 km compared to correlative measurements [Gordley et al., 1996], the agreement of MIPAS and HALOE NOx data is reasonable. Differences greater than 20% below 25 km and above 50 km are due to low NOx VMRs and are generally within the 1σ standard deviation of single scan differences.

Figure 6.

Comparison of 52 colocated nighttime MIPAS and HALOE NOx observations corrected for diurnal variations (Δ latitude < 3°, Δ longitude < 15°, Δ time < 12 hours) measured on 18–27 September and 11–13 October 2002. (left) Mean values and 1σ variances (bars) of relative differences between MIPAS and HALOE NOx versus altitude. (right) Scatterplot of the individual NOx measurements of MIPAS versus HALOE for all altitudes. Comparisons to HALOE sunrise and sunset measurements are shown in red and blue, respectively.

8. Summary and Conclusions

[41] We have described in this paper the non-LTE retrieval approach for VMRs and excitation rates of the odd nitrogen species NO and NO2 from MIPAS/Envisat limb emission spectra and its application to data recorded during July to October 2002. To our knowledge, these are the first non-LTE retrieval of stratospheric NO from atmospheric emission spectra which, in consequence, allow to infer total NOx under dark conditions (i.e., polar night). Between 20 and 60 km, the absolute precision of derived NO and NO2 VMR profiles is around 0.2–0.3 ppbv and 0.5–1.0 ppbv, respectively. This corresponds to a relative precision of retrieved NO2 of 2% at 40 km VMR peak altitude during day and 5% at 30 km VMR peak altitude during night. For daytime NO retrieval, the relative precision is around 7% at VMR peak altitude at 40 km. Major systematic error sources reducing the accuracy of retrieved NOx to 15% are uncertainties of temperature and spectroscopic data. Nitric oxide retrievals are further degraded by horizontal structures in the thermospheric NO distribution, which, depending on auroral activity, induce NO retrieval errors up to 5 ppbv for worst case conditions at polar latitudes. The vertical resolution of the retrieved NO and NO2 data is around 3.5–6.5 km at altitudes of 20–50 km, but worse (12–18 km) in the remaining altitude regions.

[42] In order to correctly consider the NO2 non-LTE effects in the retrievals, the photochemical excitation rate of NO2(v3) vibrational states was derived from NO2(002→001) emissions prior to retrievals of the NO2 VMR. The photochemical excitation rate of NO2(v3) was found to be about 50 times smaller than previously estimated from LIMS measurements. As a consequence, NO2(v3) populations keep in LTE up to approximately 50 km. At 50–60 km, the omission of non-LTE in the retrieval of NO2 abundances from the NO2(001→000) fundamental band would result in an underestimation of derived NO2 up to 30%.

[43] The NOx partitioning of the retrieved data is in excellent agreement with steady state photochemistry, which confirms the correctness of our non-LTE retrieval approach. Since self-consistency of NO and NO2 is only achieved by inclusion of stratospheric non-LTE effects in NO(v) 5.3 μm emissions, MIPAS data thus confirm the stratospheric NO(v > 0) population enhancements due to chemical excitations by NO2 photolysis as predicted by Kaye and Kumer [1987]. The disregard of these non-LTE effects in the retrieval of NO would result in an overestimation of derived NO of up to 60%. The retrieved NOx abundances are also consistent with HALOE NOx observations.


[44] The authors acknowledge ESA for providing MIPAS spectra and NILU and ECMWF for meteorological data. We thank the members of the HALOE science and data processing teams for making their data available. The IAA team was partially supported by Spanish Ministerio de Educación y Ciencia under projects REN2001–3249/CLI and ESP2004–01556. B. Funke has been supported through an European Community Marie Curie Fellowship. The IMK team was supported by the German HGF-Vernetzungsfonds ‘Envisat’ (BMBF 01SF9953/8), KODYACS (BMBF 07ATF43), SACADA (BMBF 07ATF53), and by the EU-Projects AMIL2DA (EVG1-CT-1999-00015) and TOPOZ-III (EVK2-CT-2001-00102).