The cryogenic Fourier transform limb emission spectrometer GLORIA operates in the thermal infrared spectral region between about 7 and 13 µm using a two-dimensional detector array observing 128 vertical and 48 horizontal interferograms per measurement (Friedl-Vallon et al., 2014; Riese et al., 2014, and references therein). The interferograms are Fourier transformed into the spectral domain and radiometrically calibrated using in-flight measurements of two black bodies (Kleinert et al., 2014). Furthermore, spectra in the horizontal direction were averaged to improve the signal-to-noise ratio such that the noise equivalent spectral radiance (NESR) finally amounts to about 1.5×10-8 W (cm2 sr cm-1)-1. Spectra recorded with maximum optical path difference of 8.0 cm, which corresponds to an un-apodized spectral resolution of 0.0625 cm-1, were used for this study. These so-called chemistry mode spectra are apodized with the Norton and Beer (1976) “strong” function. Due to the high spectral resolution, these measurements allow for the retrieval of many species with minor contribution to the spectra by the separation of individual spectral lines from continuum-like emissions. A comprehensive description of technical details of the GLORIA instrument is given by Friedl-Vallon et al. (2014) and Riese et al. (2014).

In this study, we report results from the WISE aircraft campaign, which was dedicated mainly to the investigation of dynamical processes and the evolution of air masses within the context of stratosphere–troposphere exchange. Sixteen flights with HALO were performed from Shannon (Ireland) and Oberpfaffenhofen (Germany) between 31 August and 21 October 2017, in a region within about 40∘ W and 15∘ E longitude and 40 and 75∘ N latitude. In the following sections, we focus on results of the flight on 13 September 2017 where strong enhancements in the VMR of the pollutant species C2H6, C2H2, HCOOH, and PAN were observed by GLORIA. Figure 1 shows the flight path consisting of two main legs together with tangent points of GLORIA where the instrument was operated in the chemistry mode.

Radiances are calculated with the Karlsruhe Optimized and Precise Radiative transfer Algorithm (KOPRA; Stiller et al., 2002) based on spectroscopic parameters from the high-resolution transmission molecular absorption database (HITRAN; Gordon et al., 2017). The retrieval itself is performed with the procedure KOPRAFIT (Höpfner et al., 2002) using derivatives of the radiance spectrum with respect to atmospheric state and instrument parameters (Jacobians) calculated by KOPRA. The inverse problem of radiative transfer is solved by KOPRAFIT with a Gauss–Newton iterative method (Rodgers, 2000) in combination with a Tikhonov–Phillips regularization approach (Phillips, 1962; Tikhonov, 1963) using a constraint with respect to a first derivative of the a priori profile of the target species. A detailed description of the general retrieval process together with the validation of major species (e.g. O3, HNO3, and ClONO2) observed by GLORIA is given by Johansson et al. (2018).

In a first step cloud-affected spectra are filtered out using a cloud index as described in Spang et al. (2004). Spectra with a colour ratio of the mean radiance in two spectral windows (788.20–796.25 and 832.30–834.40 cm-1) larger than 2 were chosen for the retrieval process. Prior to the trace gas retrievals, the pointing elevation angle was retrieved to compensate for systematic misalignment of the line of sight (LOS) of the instrument according to the method described in Johansson et al. (2018). The final pointing (LOS) error was estimated to 0.01∘. Vertical profiles of atmospheric species were taken from a climatological atmosphere (Remedios et al., 2007), updated with surface concentration data from NOAA ESRL GMD (National Oceanic and Atmospheric Administration, Earth System Research Laboratory, Global Monitoring Division; Montzka et al., 1999). Subsequently, a temperature retrieval was carried out using pressure–temperature a priori data from the European Centre for Medium-Range Weather Forecasts (ECMWF), which was interpolated to the GLORIA vertical retrieval grid. Spectral windows around 811 and 957 cm-1 containing appropriate CO2 transitions were used for this analysis. The total error of the temperature retrieval calculations is estimated to about 1.5 K (Johansson et al., 2018).

The spectral analysis of the target species C2H6, C2H2, HCOOH, and PAN is impeded by overlapping features of so-called disturbing gases in the corresponding spectral region. Hence, the retrieval of these pollution trace gases is more challenging compared to the species discussed in Johansson et al. (2018). A careful selection of appropriate microwindows is essential to perform retrievals of these pollutants with good accuracy. Test retrievals were used to identify microwindows that combine limited overlap of spectral signatures of disturbing gases with a high sensitivity to changes in the abundance of target gases. Main interfering species were either adjusted simultaneously together with the target molecule or pre-fitted in a different spectral interval.

The analysis of C2H6 was performed in four microwindows within the ν9 band between 819 and 833 cm-1 (see Fig. 2a). Many overlapping features of so-called disturbing gases are visible in this spectral region: first of all H2O and CO2. The error budget of C2H6 is shown in Fig. 2b. At higher altitudes with low C2H6 values, the total error of this molecule is dominated by the temperature error. Lower down in the VMR maximum region, the systematic spectroscopic error of C2H6 (as given in HITRAN) governs the error budget. The total error of C2H6 remains within about 15 % in the altitude region of the upper troposphere.

Retrieval calculations of C2H2 were carried out in the R-branch region of the ν5 band. Figure 3a shows spectral contributions of relevant species in four microwindows between 759 and 781 cm-1 that have been found most appropriate to derive C2H2 from GLORIA spectra. The corresponding retrieval error budget of C2H2 is given in Fig. 3b. The random noise error is dominating the budget over nearly the complete altitude range. The total error of C2H2 stays within 10 %–15 % in the region of the VMR maximum in the upper troposphere.

The analysis of the molecule HCOOH was performed in the spectral range between 1086 and 1117 cm-1 (see Fig. 4a). Three microwindows were chosen including the strong Q-branch of the HCOOH ν6 vibrational band. The spectral region is dominated by spectral features due to O3, CO2, CFC-12, and HCFC-22 transitions. The spectroscopic part of the total error is dominant in the altitude region of the HCOOH VMR maximum in the upper troposphere (see Fig. 4b). Here, the total HCOOH error stays within 10 %. Apart from this altitude region, the error budget is governed by the random noise part and the total error increases significantly (mainly in the upper part of the profile with low HCOOH values).

The retrieval of PAN was conducted in two broad microwindows of the ν16 band between 780 and 805 cm-1 (see Fig. 5a). The spectral gap between these microwindows was chosen such that the strong CO2 Q-branch at 792 cm-1 was excluded from the data analysis. Besides CO2, the molecules H2O and CCl4 are the main interfering species in the spectral region of PAN. The retrieval error budget is depicted in Fig. 5b. Besides random noise, further error sources like inaccuracies in the FOV and the retrieved temperature profile contribute to the total PAN error that remains within 10 % in altitude regions with enhanced PAN amounts.

Besides the retrieval of the above-mentioned pollutant gases, the tracer species ozone was also inferred from the recorded spectra. Many spectral ozone lines are available in the mid-infrared spectral region. Transitions between 780 and 788 cm-1 within the ν2 band were chosen for the retrieval process similar to the method described in Johansson et al. (2018). The total ozone error is within 10 % with a vertical resolution of 0.3 to 1.5 km.

The altitude resolution of all retrievals, calculated from the full width at half maximum of the rows of the averaging kernel matrix, was used as an a posteriori quality filtering of the retrieved data. Only vertical profile parts with an altitude resolution of better than 2 km were finally used for the data interpretation.

Retrieved volume mixing ratios of C2H6, C2H2, HCOOH, and PAN together with O3 inferred from limb emission spectra during the WISE flight on 13 September 2017 are displayed in Fig. 6. Ozone is a molecule with highest concentrations in the stratosphere (Brasseur and Solomon, 2005). Hence, it can be used as a tracer to diagnose detected air masses whether they are of stratospheric or tropospheric origin. The general shape of O3 VMR is strongly correlated with the tropopause as shown in Fig. 6a. There are two regions where stratospheric air comes down to about 7 km (around 14:45 and around 16:50 UTC). Here, the tropopause layer reaches these low altitudes in the form of a stratospheric intrusion, while at the beginning and the end of the measurement phase the troposphere extends up to about 12 km. Trajectory calculations have shown that, due to a west-southwesterly mid- and upper tropospheric air flow in the region of the flight path shown in Fig. 1, GLORIA has sounded virtually the same air mass twice, yielding to a kind of symmetry in the horizontal trace gas distribution before and after 16:10 UTC. Measured stratospheric ozone volume mixing ratios are within 0.1 and 0.8 ppmv. These values are in line with other mid-latitude remote sensing and in situ observations performed in this altitude range (e.g. Cortesi et al., 2007; Livesey et al., 2008; Bourgeois et al., 2020).

Measured concentrations of the species C2H6 are shown in Fig. 6b. The VMR distribution of C2H6 is in parts anti-correlated to the one of ozone. Two regions of stronger enhanced VMR up to about 2.2 ppbv can be seen in the upper troposphere at the beginning and at the end of the measurement period over the region south of Ireland and near the coastline of the Netherlands and Belgium. In the stratosphere, no stronger enhanced C2H6 VMR levels are visible and values remain below about 0.6 ppbv, which can be confirmed by mid-latitude satellite observations (Rinsland et al., 2005; Glatthor et al., 2009; Wiegele et al., 2012). For instance, up to about 1.7 ppbv C2H6 was observed by the Atmospheric Chemistry Experiment (ACE) instrument in the mid-latitude upper troposphere, but with less spatial resolution compared to GLORIA. In addition, in situ aircraft measurements recorded C2H6 amounts of more than 2 ppbv in the mid-latitude upper troposphere (Rudolph, 1995; Monks et al., 2018, and references therein) in accordance with the high values seen by GLORIA.

The same behaviour as seen for C2H6 is present in the vertical and horizontal distribution of C2H2 amounts (see Fig. 6c). Elevated volume mixing ratios of up to 0.2 ppbv in the upper troposphere are clearly visible at the beginning and close to the end of the observation period. In the stratosphere, measured C2H2 values appear noisy and stay clearly below 0.1 ppbv most of the time. C2H2 VMR measured by GLORIA lies within the range of satellite (Rinsland et al., 2005; Wiegele et al., 2012) and aircraft in situ data (Xiao et al., 2007, and references therein) obtained in the same altitude region at mid-latitudes with values of up to 0.45 ppbv as seen by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) instrument under biomass burning conditions in October 2007 (Wiegele et al., 2012).

Measured HCOOH volume mixing ratios are depicted in Fig. 6d. As in the case of the previously mentioned species C2H6 and C2H2, large amounts of HCOOH are also visible at high altitudes in the troposphere during early and late times of the GLORIA observations with values up to 0.9 ppbv. Comparable amounts were seen by airborne in situ measurements in the mid-latitude upper troposphere (Reiner et al., 1999; Singh et al., 2000) with VMR maxima of up to 0.6 ppbv. Very high upper-tropospheric HCOOH mixing ratios of 3.1 ppbv and 2.0 ppbv were observed by the space-borne ACE Fourier transform spectrometer in two preceding years in austral spring (Rinsland et al., 2006). In contrast, stratospheric HCOOH values are low and not higher than 0.1 ppbv. These values are in accordance with space-borne mid-latitude observations by the ACE Fourier transform spectrometer and the MIPAS instrument in the altitude regime considered here (Rinsland et al., 2006; Grutter et al., 2010).

The two-dimensional cross section of PAN is shown in Fig. 6e. The distribution of VMR maxima and minima is more structured compared to the one of the previously regarded pollution trace gases. Increased amounts of PAN up to 0.4 ppbv are not only visible at the beginning and end of the observation period in the upper troposphere but also in the lowermost stratosphere around 15:00 UTC, at altitudes of 7 to 8 km. Somewhat less enhanced quantities are noticeable near 14 km around 15:30 UTC. Elevated PAN amounts of comparable magnitude have also been detected in the UTLS region by space-borne instruments. The Cryogenic Infrared Spectrometers and Telescopes for the Atmosphere (CRISTA) experiment detected up to 0.35 ppbv PAN in the mid-latitude UTLS over East Asia and the western Pacific (Ungermann et al., 2016). The ACE Fourier transform spectrometer recorded 0.52 ppbv PAN at 11.5 km in a biomass burning plume above eastern Africa in autumn 2005 (Coheur et al., 2007) and up to 0.6 ppbv PAN were seen by the MIPAS instrument under biomass burning conditions in the upper troposphere in October 2007 (Wiegele et al., 2012). PAN mixing ratios of up to 0.2 ppbv were detected by airborne in situ measurements in the free troposphere (Singh et al., 2000, 2001). A compilation of different aircraft observations given by Fischer et al. (2014) exhibits mid-latitude middle and upper-tropospheric PAN values of more than 0.5 ppbv over the Northern Atlantic during summer. These different shapes of VMR distributions might be explained by very long lifetimes of PAN under cold UTLS conditions and by the fact that PAN is not emitted directly but dependent on the availability of precursor substances as described in Sect. 1.

The comparison of measured species to model simulations is presented in Figs. 7 and 8. The observed data have been temporally smoothed with a 39-point adjacent averaging routine to permit a more realistic comparison given the different horizontal resolutions of the measurement and the EMAC and CAMS simulations. For O3, both simulations generally reproduce the tropospheric and stratospheric concentrations seen by GLORIA (see Fig. 7a–c) but with coarser spatial structure. The chemistry climate model EMAC is able to simulate finer structures while CAMS only produces a smooth distribution of assimilated ozone. Both models tend to slightly overestimate the amount of ozone in the troposphere. This is also visible in Fig. 8a–c where differences between both EMAC runs (with and without enhanced NMVOC emissions) and GLORIA observations are shown. The amount of simulated ozone in the EMAC_2 run is only slightly higher (less than 10 ppbv) compared to the EMAC simulation without enhanced NMVOC emissions (VMR differences in Fig. 8b and c are therefore nearly the same).

Comparisons for C2H6 are displayed in Fig. 7d–f. Both models are able to qualitatively reproduce the temporal and spatial region of enhanced upper-tropospheric C2H6 as observed by GLORIA. As in the case of ozone, EMAC again is able to display finer structures in the vertical and horizontal distribution of C2H6 compared to CAMS. However, deficits in the simulated absolute C2H6 quantities are clearly visible in both models, especially in the case of CAMS. A considerable underestimation of CAMS C2H6 with respect to airborne observations was already reported by Wang et al. (2020). The EMAC_2 simulation with increased NMVOC emissions at least reduces the difference to the GLORIA observations compared to the EMAC run without these stronger NMVOC emissions (see Fig. 8d–f).

For C2H2, we note that EMAC predicts elevated concentrations in much the same region where GLORIA reports enhancements (see Fig. 7g–h). In addition, the measured VMRs in the upper troposphere are only a little underestimated in terms of their absolute amount by the EMAC_2 simulation using raised NMVOC emissions. In the stratosphere, simulated C2H2 amounts are too low compared to the measurement. Using standard NMVOC emissions in EMAC leads to an increased underestimation of C2H2 amounts compared to GLORIA (see Fig. 8g–i).

The comparison of the species HCOOH is shown in Figs. 7i–k and 8j–l. Elevated HCOOH concentrations, as recorded by GLORIA in the upper troposphere, are clearly underestimated by both models, especially in the CAMS simulation, although the atmospheric region of the (too weak) enhanced HCOOH amounts in the models agrees with the measured one. However, the EMAC_2 simulation at least reduces differences with respect to the GLORIA observations.

Looking at the temporal and spatial distribution of PAN the situation appears somewhat different to the comparisons discussed above (see Figs. 7l–n and 8m–o). The principal behaviour of enhanced PAN values in the upper troposphere is captured by both atmospheric models. EMAC produces slightly finer structures in the stratosphere compared to CAMS. However, the measured small-scale variations in the amount of PAN especially near 14:30 UTC between 6 and 8 km are not reproduced by the EMAC_2 simulation with enhanced NMVOC emissions, while the observed elevated PAN values around 15:00 UTC in the lowermost stratosphere from 7 to 8 km are also visible in the model output. Apart from the regions with the highest measured PAN amounts, EMAC_2 tends to overestimate the concentration of PAN below about 13 km (what is not the case in the standard EMAC run). Interesting VMR variations are also seen by GLORIA in the stratosphere above 13 km. The PAN VMR maximum detected around 15:40 UTC near 14 km is not visible in the model simulations. The PAN VMR minima near 15:00 UTC and between 16:00 and 16:15 UTC at about 13.5 km are reflected as a VMR minimum in EMAC_2, although with lower absolute quantities. The different shapes of the horizontal and vertical distributions of PAN VMR is most probably caused by long-range atmospheric transport. Compared to the species discussed before, sources and sinks are different, and atmospheric lifetime of PAN is considerably longer. In the following subsection, we will focus on the origin of the polluted air masses, which have been detected by GLORIA.

To estimate the geographical region (within the Earth's upper planetary boundary layer) of the origin of the measured enhanced amounts of the pollutants and their pathways, we performed backward trajectory calculations as well as global three-dimensional CLaMS simulations with artificial tracers of air mass origin as described in the following subsections.