MIPAS IMK / IAA Carbon Tetrachloride ( CCl 4 ) Retrieval and First Comparison With Other Instruments

MIPAS thermal limb emission measurements were used to derive vertically resolved profiles of carbon tetrachloride (CCl4). Level-1b data versions MIPAS/5.02 to MIPAS/5.06 were converted into volume mixing ratio profiles using the level-2 processor developed at Karlsruhe Institute of Technology (KIT) Institute of Meteorology and Climate Research (IMK) and Consejo Superior de Investigaciones Científicas (CSIC), Instituto de Astrofísica de Andalucía (IAA). Consideration of peroxyacetyl nitrate (PAN) as an interfering species, which is jointly retrieved, and CO2 line mixing is crucial for reliable retrievals. Parts of the CO2 Q-branch region that overlap with the CCl4 signature were omitted, since large residuals were still found even though line mixing was considered in the forward model. However, the omitted spectral region could be narrowed noticeably when line mixing was accounted for. A new CCl4 spectroscopic data set leads to slightly smaller CCl4 volume mixing ratios. In general, latitude–altitude cross sections show the expected CCl4 features with highest values of around 90 pptv at altitudes at and below the tropical tropopause and values decreasing with altitude and latitude due to stratospheric decomposition. Other patterns, such as subsidence in the polar vortex during winter and early spring, are also visible in the distributions. The decline in CCl4 abundance during the MIPAS Envisat measurement period (July 2002 to April 2012) is clearly reflected in the altitude–latitude cross section of trends estimated from the entire retrieved data set.


Introduction
Carbon tetrachloride (CCl 4 ) is an anthropogenically produced halogen-yielding trace gas and partly responsible for stratospheric ozone depletion.It is also a potent greenhouse gas with a 100-year global warming potential of 1730 (IPCC, 2013;World Meteorological Organization, 2014).CCl 4 was commonly used in fire extinguishers, as a precursor to refrigerants, and in dry cleaning prior to 1990, when it was restricted within the framework of the London Amendment to the Montreal Protocol.Its abundances in the atmosphere increased steadily from the first part of the 20th century.Emissions declined significantly after 1990, as did the amount of CCl 4 in the atmosphere with a few years delay.2007-2012 bottom-up emissions of 1-4 kilotonnes year −1 were assessed by combining country-by-country reports to the United Nations Environmental Programme (UNEP) (Liang et al., 2016).This bottom-up estimate differs considerably from the 57 (± 17) kilotonnes year −1 top-down emissions Published by Copernicus Publications on behalf of the European Geosciences Union.
which were evaluated in 2014 (World Meteorological Organization, 2014) using atmospheric measurements and lifetime estimates.Even when possible CCl 4 precursors and unreported, inadvertent emissions are accounted for, the gap between reported bottom-up and estimated top-down CCl 4 emissions cannot be closed, as bottom-up emissions still only add up to 25 kilotonnes year −1 (SPARC, 2016).Besides a sink in the atmosphere, CCl 4 is decomposed in the ocean and the soil with different lifetimes for each sink.Reassessment of the different lifetime estimates, which are essential for an adequate top-down assessment of emissions, leads to lower emissions of ∼ 40 (± 15) kilotonnes year −1 .While the gap between bottom-up and top-down emissions is now smaller after reassessments, the discrepancy is still not solved entirely.Previous measurements of stratospheric CCl 4 have also been performed by the Atmospheric Chemistry Experiment Fourier Transform Spectrometer (ACE-FTS), a cryosampler instrument employed at Frankfurt University, and the balloon-borne version of the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS-B2).The first version of the balloon-borne MIPAS instrument (MIPAS-B) and ATMOS (Atmospheric Trace Molecule Spectroscopy) also measured CCl 4 , but not during the MIPAS Envisat measurement period (Zander et al., 1996;von Clarmann et al., 1995).Additional measurements, especially vertically wellresolved ones with global coverage such as satellite measurements from MIPAS, can help to improve the understanding of the atmospheric CCl 4 budget and stratospheric lifetime estimate.Furthermore, as a tracer with relatively long stratospheric lifetimes, CCl 4 measurements can improve the understanding of changes in the Brewer-Dobson circulation by further constraining the lower boundary, e.g.processes around the tropopause.In this study, we present the retrieval of CCl 4 distributions from MIPAS limb emission spectra.First, we characterize the MIPAS instrument (Sect.2), followed by a detailed description of the retrieval and the specific issues that had to be dealt with to derive CCl 4 concentrations (Sect.3).We then compare the results of the MIPAS Envisat CCl 4 retrieval with those of ACE-FTS, those of the second balloon-borne MIPAS instrument (MIPAS-B2) and those of cryosampler measurements (Sect.5) and summarize the results in the conclusions (Sect.6).

MIPAS
The Michelson Interferometer for Passive Atmospheric Sounding was one of the instruments aboard the European Environmental Satellite (Envisat).It was launched into a sunsynchronous orbit at an altitude of approximately 800 km on 1 March 2002.On 8 April 2012, all communication with the satellite was lost, ending an observation period of more than 10 years.Envisat orbited the earth 14.4 times a day, crossing the Equator at 10:00 and 22:00 local time.MIPAS measured infrared emissions between 685 and 2410 cm −1 (14.6 and 4.15 µm) (Fischer et al., 2008), which allows for day and nighttime measurements with global coverage.The initial spectral resolution of the instrument was 0.025 cm −1 (0.0483 cm −1 after a "Norton-Beer strong" apodization; Norton and Beer, 1976).An instrument failure in March 2004 led to an observation gap until January 2005 when the instrument was successfully restarted.The first period (June 2002to March 2004) is referred to as full spectral resolution (FR) period, while the period from January 2005 to April 2012 is referred to as reduced spectral resolution (RR) period.Due to a problem with one of the interferometer slides, MIPAS could only be operated with a spectral resolution of 0.0625 cm −1 (0.121 cm −1 apodized) from January 2005 on.In this study, only measurements from the instrument's "nominal operation mode" are used.In this mode, the number of tangent altitudes increased from 17 during the FR period to 27 during the RR period.The vertical coverage ranges from 6 km to around 68 km during the FR period and up to around 70 km during the RR period, respectively.MIPAS initially took around 1000 measurements per day.In 2005, operation was resumed at reduced duty cycle.By the end of 2007, MIPAS was back at full duty cycle, which amounts to approximately 1300 RR measurements per day.The horizontal sampling changed from 510 km during the FR period to 410 km during the RR period.
The temperature and various atmospheric trace gases are retrieved from level-1b data using a retrieval processor developed at the Institute of Meteorology and Climate Research at the Karlsruhe Institute of Technology (KIT) in close cooperation with the Instituto de Astrofísica de Andalucía (CSIC) in Granada, Spain.Results shown in this publication cover both the FR and the RR period.

Retrieval
The MIPAS Envisat retrieval is based on a non-linear leastsquares approach and employs a first-order Tikhonov-type regularization (von Clarmann et al., 2003(von Clarmann et al., , 2009)).
The radiative transfer is modelled using the Karlsruhe Optimized and Precise Radiative Transfer Algorithm (KOPRA) model (Stiller, 2000).The spectral regions used for the retrieval of CCl 4 are 772.0-791.0 and 792.0-805.0cm −1 .The gap from 791.0 to 792.0 cm −1 is necessary, since even when accounting for line mixing, strong effects from the CO 2 Qbranch still occur in the residuals.Several results from previous steps in the retrieval chain were used to derive CCl 4 (Table 1) including the spectral shift (z tangent ), the temperature (T ), the horizontal temperature gradient (T grad ) and mixing ratio profiles of HNO 3 , ClO, CFC-11, C 2 H 6 , HCN, ClONO 2 and HNO 4 .In addition, several species were found to improve the retrieval whenever their mixing ratio profiles were fitted alongside CCl 4 .These are peroxyacetyl nitrate (PAN), CH 3 CCl 3 , HCFC-22, O 3 , H 2 O, C 2 H 2 and COF 2 .Although for most of these species results from preceding retrieval Atmos.Meas. Tech., 10, 2727-2743, 2017 www.atmos-meas-tech.net/10/2727/2017/Table 1.Retrieval details on the spectroscopic region, species imported from preceding retrieval steps and variables fitted jointly during the retrieval process.Brackets denote mixing ratios.

Spectral regions
Imported from preceding Jointly fitted retrieval steps variables 772.0-791.0cm −1 Shift(z tangent ) [PAN](z) 792.0-805.0cm −1 T (z) steps are available, fitting their concentrations jointly with that of CCl 4 reduces the fit residuals significantly.This is attributed to spectroscopic inconsistencies of the interferers' spectroscopic data between the spectral region where these were retrieved and the spectral region where CCl 4 is analyzed.Also fitted were a background continuum accounting for spectral contributions from aerosols and a radiance offset which is constant for all tangent altitudes (Table 1).These retrieval settings lead to spectral fits as displayed in Figs. 1  and 2, where an example for the FR period and the RR period is shown, respectively.The measured spectra are plotted in black (not discernible from the best fit modelled in the fitting window), while the red and the blue lines represent the modelled spectra of the regions from 772.0 to 791.0 and from 792.0 to 805.0 cm −1 , respectively.Some periodic residuals are visible in both the FR and the RR period.These result from less than perfectly fitted CO 2 but, as will be shown in Sect.5, are only of minor relevance for the accuracy of the retrieved CCl 4 .

Information cross-talk with PAN
The signature of PAN is particularly prominent in the spectral region of CCl 4 and can thus be retrieved during the same retrieval step.Actually, jointly fitting PAN is very important for the CCl 4 retrieval.Since PAN was already retrieved from MIPAS spectra before (Glatthor et al., 2007), it is of obvious interest to investigate the PAN results from the CCl 4 -PAN joint retrieval in comparison with those from the original PAN retrieval.There, CCl 4 was fitted alongside PAN but the retrieval was not yet optimized for CCl 4 .
We find slightly higher volume mixing ratios of PAN throughout most of the altitude-latitude cross section (Fig. 3).As a consequence, areas showing unphysical mixing ratios below zero in the original retrievals (left panel of Fig. 3) are now slightly positive or very close to zero.This suggests that jointly fit PAN from the retrieval optimized for CCl 4 might be more accurate than PAN retrieved using the old CCl 4 distributions.

Line mixing
Since the spectral region where CCl 4 is retrievable contains a CO 2 Q-branch, the retrieval is set up to account for   line mixing (Funke et al., 1998).This was done by using the Rosenkranz approximation (Rosenkranz, 1975).Tests were also performed using the computationally more demanding direct diagonalization, but this approach was not found to noticeably change the results of the retrieval.This is possibly the case because the microwindows were carefully selected to omit major spectral signatures of the CO 2 Q-branch and because the effect of line mixing is generally smaller at stratospheric pressure levels.However, it was still necessary to omit parts of the CO 2 Q-branch.Figures 4 and 5 show spectra where the full spectral region was fitted.In Fig. 4, line mixing was not considered and thus a large peak in the residual is visible close to 791.0 cm −1 .In Fig. 5, the Rosenkranz approximation was used to account for line mixing.Even though the residual is considerably smaller than without line mixing taken into account -as would be expected -peaks significantly larger than for the remainder of the window are still visible between 791.0 and 792.0 cm −1 .Although inclusion of line mixing significantly reduces the residuals in the CO 2 branch, the residuals are still unacceptably large there.With the Rosenkranz approximation, however, the spectral region excluded from the fit could be narrowed from 791.0 to 792.0 cm −1 from 790.5 to 792.5 cm −1 .

New CCl 4 spectroscopic data
During the ongoing development of the MIPAS Envisat CCl 4 retrieval, a new CCl 4 spectroscopic data set was published by Harrison et al. (2017).which, in the tropopause region, agree better with tropospheric measurements.Tropospheric volume mixing ratios are reported to be at approximately 95 pptv, which is very close to what MIPAS Envisat presents around the tropical tropopause and at midlatitudes of the Northern Hemisphere when using the new spectroscopic data set.In contrast, using HITRAN 2000 sometimes results in volume mixing ratios above 100 pptv in the same region.Thus, we consider the new spectroscopic data set more adequate for the retrieval of CCl 4 .

Distributions
Figures 7 and 8 and the lower panel of Fig. 6 give an overview of the latitudinal and altitude distribution of CCl 4 of different time periods.All of the altitude-latitude cross sections show the expected patterns of CCl 4 with a rapid decrease with increasing altitude in the stratosphere, as the gas is photolyzed there.In addition, highest volume mixing ratios appear at the Equator where CCl 4 , along with many other trace gases, enters the stratosphere due to the upward transport associated with the Brewer-Dobson circulation.During January 2010, March 2011 and particularly April 2011, subsidence of higher stratospheric air results in reduced mixing ratios over the North Pole.In Spring 2011, an unusually stable northern polar vortex resulted in severe ozone depletion and particularly strong subsidence (Manney et al., 2011;Sinnhuber et al., 2011), which is reflected in the observations shown here.In general, MIPAS Envisat shows higher volume mixing ratios in the lower stratosphere during the FR period, which fits well with the overall decline in CCl 4 abundance in the atmosphere due to its restriction under the Montreal Protocol.This impression is also supported by the lower panel in Fig. 6, which shows lower overall volume mixing ratios than MIPAS sees during the FR period but which are still slightly higher than during 2010 and 2011.All cross sections show a maximum in the CCl 4 volume mixing ratios around the tropical tropopause connected with values of similar magnitude at lower altitudes of northern extratropical regions.This pattern was also seen in HCFC-22 (Chirkov et al., 2016) and could be linked to the Asian monsoon.Calculations with the Chemical Lagrangian Model of the Stratosphere (CLaMS) by Vogel et al. (2016) show that there indeed exists a mechanism which can produce local maxima in the upper troposphere in 2-D distributions of source gases.So, the monsoon might offer an explanation for the patterns seen in CCl 4 around these atmospheric regions as well.

Altitude resolution
The vertical resolution of the CCl 4 profiles is very similar for the FR and the RR period.From about 2.5 to 3 km at the lower end of the profiles, it degrades to approximately 5 km at ∼ 25 km and ∼ 7 km at ∼ 30 km, calculated as the full width at half maximum of the row of the averaging kernel matrix (Rodgers, 2000).The degrees of freedom are usually around 3.5 for the FR period and close to 4.0 for the RR Atmos.Meas.Tech., 10, 2727-2743, 2017 www.atmos-meas-tech.net/10/2727/2017/period (Fig. 9).This is presumably attributed to the finer vertical sampling during the RR period with 27 tangent altitudes compared to 17 tangent altitudes during the FR period.The signal decreases rapidly with altitude, as the volume mixing ratios of CCl 4 do.Above 30 km, hardly any CCl 4 information is available in the MIPAS spectra.Slightly below 20 km, the averaging kernels show negative side wiggles which are more pronounced during the FR period (left panel) than the RR period (right panel).

Error budget
Tables 2 and 3 list the error budgets for midlatitudes during the FR and RR period between 10 and 40 km.Examples for other latitudes can be found in the Appendix (Tables A1-A6).For legibility reasons, the errors are only given every 5 km, although the retrieval grid is 1 km.Errors due to elevation uncertainties of the line of sight (LOS) and uncertainties of several contributing species are given.All profiles show a strong increase in the relative errors at and above 30 km.During the FR period, the absolute total errors are fairly similar below this altitude, while large differences can occur from 20 km upwards.Absolute errors are close to 3 pptv between 10 and 25 km and around 5 to 6 pptv at 15 km where larger error appear for all atmospheric situations except the polar summer one where the errors stay close to 3 pptv.The largest error component is measurement noise (third column), while at 15 km significant parameter errors have to be considered, in particular the elevation uncertainties of the LOS and instrument line shape (ILS).Beyond this, uncertainties of HNO 4 and ClONO 2 profiles, frequency calibration (shift) and temperature contribute to the total error.The decrease of retrieval noise towards higher altitudes is explained by the coarser altitude resolution at higher altitudes.For the RR period, the patterns look slightly different.There is no peak in the total error around 15 km, but the total error is either rather constant at lower altitudes or decreases with altitude.Contributions to the error budget are, however, similar to the FR period.
Figure 10 compares the estimated total error with the deviation of the profiles in a quiescent atmosphere.This comparison was created in a similar way as in Eckert et al. (2016, Sect. 6).Up to 18 km altitude, the sample standard deviation of MIPAS Envisat results is only slightly larger than the estimated error.Thus, these profiles suggest that the estimated error can explain most of the variability in the CCl 4 profiles up to approximately 18 km.Correspondingly, the error estimate can be considered realistic from the bottom of the profile up to this altitude.timates are the underlying MIPAS spectra.We use MIPAS V5 spectra which were found to be subject to an instrument drift due to detector aging (Eckert et al., 2014).Valeri et al. (2017) use version 7 spectra, where an attempt was made to tackle the problem of detector aging during the level-1 processing.However, Hubert et al. (2016) show that there is still a drift problem in the version 7 MIPAS temperatures.Since these temperature drifts are expected to propagate onto the retrieved CCl 4 mixing ratios, it is not clear if version 5 or version 7 is more adequate for trend analysis.In spite of these differences and technical differences in the level-2 data processing, the trends inferred by Valeri et al. ( 2017) and ours show important common features.In both data sets a hemispheric asymmetry is found, with   The ATMOS instrument measured in solar occultation covering the spectral region from 600 to 4700 cm −1 with a spectral resolution of 0.01 cm −1 .ATMOS took measurements in 1985, 1992, 1993 and 1994.The ATMOS profiles shown in Fig. 12 were extracted directly from Zander et al. (1996, Fig. 1).CCl 4 volume mixing ratio profiles in the subtropics (20-35 • N; thin dashed lines) and at midlatitudes (35-49 • N; thin full lines) are presented there.Measurements were taken from 3 to 12 November in 1994 during the ATLAS-3 shuttle mission.We depicted midlatitude profiles as solid lines and subtropical profiles as dashed lines in Fig. 12 of this paper.To compare the ATMOS profiles with MIPAS Envisat, we used MIPAS Envisat data of all years from 3 to 12 November and calculated an arithmetic mean for both latitude bands (subtropics and midlatitudes).In Fig. 12, MIPAS Envisat profiles are shown in blue, while the ATMOS profiles are shown in orange.The ATMOS profiles show higher volume mixing ratios than those of MIPAS Envisat, because they were measured shortly after CCl 4 emissions were restricted and, thus, volume mixing ratios used to be higher in the atmosphere.However, the general shapes of the ATMOS profiles agree well with those of MIPAS Envisat.Both MIPAS Envisat and ATMOS show CCl 4 mixing ratios which quickly decrease with altitude.The slopes of decline are similar above ∼ 20 km.Largest differences are visible at the lower end of  2) for the time around and shortly after 1990.Taking the temporal development of the surface mixing ratios into account, ATMOS and MIPAS Envisat measurements provide a coherent picture.

MIPAS-B
The first balloon-borne version of the MIPAS instrument was developed prior to the satellite instrument in the late 1980s and early 1990s at the Institute of Meteorology and Climate Research (IMK) in Karlsruhe (Fischer and Oelhaf, 1996).Measurements with this instrument have been taken since 1989 (von Clarmann et al., 1993) and first profiles of CCl 4 were derived from a flight at Kiruna, Sweden, on 14 March 1992 (von Clarmann et al., 1995).Due to the strong decrease of CCl 4 with altitude, a clear signal of the gas could not be identified at tangent altitudes of 14.5 km and above.Thus, only the spectrum at 11.3 km was analyzed and the total amount of CCl 4 was estimated by scaling the vertical profile and using information on the shape as measured in polar winter conditions before.This leads to an estimated concentration of approximately 110 pptv at 11.3 km, which is slightly higher than the peak surface values in the long time series of CCl 4 shown in Liang et al. (2016).Groundbased measurements shown in there support favouring the MIPAS Envisat CCl 4 retrieval with the new spectroscopic data set, since respective results agree better with measurements shown in Liang et al. (2016).MIPAS-B results overestimate the ground-based measurements slightly providing a consistent picture when taking differences in the volume mixing ratios into account which result from the old versus the new spectroscopic data set.

Comparisons with collocated measurements
All collocated measurements were analyzed using spectroscopic data of Nemtchinov and Varanasi (2003), which are included in the HITRAN 2000 database (Rothman et al., 2003).Thus, in order to allow for a meaningful comparison and not to mask possible other differences, a dedicated MIPAS Envisat comparison data set was generated which is based on these spectroscopic data as well.

ACE-FTS
The Atmospheric Chemistry Experiment Fourier Transform Spectrometer is one of two instruments aboard the Canadian Satellite SCISAT-1.On 12 August 2003, it was launched into a 74 • orbit at 650 km to ensure a focus on higher latitudes.It covers the globe from 85 • S to 85 • N. Since ACE-FTS is an occultation instrument, it takes measurements during 15 sunrises and 15 sunsets a day within two latitude bands.The vertical scan range covers altitudes from the middle troposphere up to 150 km.Wavelengths between 750 and 4400 cm −1 (13.3 and 2.3 µm) can be detected with a spectral resolution of 0.02 cm −1 .The vertical sampling depends on the altitude as well as the beta angle.The latter is the angle between the orbit track and the path from the instrument to the sun.The sampling ranges from ∼ 1 km between 10 and 20 km to ∼ 2-3.5 km around 35 km and declines to 5-6 km at the upper end of the vertical range.The field of view covers 3-4 km, which is approximately similar to the vertical resolution of the instrument.Comparisons in this study were made using version 3.5 of the ACE-FTS data.The CCl 4 retrieval is performed between 787.5 and 805.5 cm −1 at altitudes from 7 to 25 km (Allen et al., 2009).
For the comparison with ACE-FTS (Fig. 13), coincident profiles within 2 h time difference and no further than 5 • latitude and 10 • longitude away were used.Profiles at latitudes higher than 60 • S were omitted.Between the lower end and ∼ 16 km the agreement is always close to 10 %, with slightly larger differences below 10 km than between 10 and 15 km.Above 15 km, the mean profiles deviate more strongly and exceed relative differences of 50 % above 19 km (Fig. 13d).However, differences above 19 km are not as apparent in the absolute comparison (Fig. 13a).The volume mixing ratio difference stays within similar values up to near 25 km.Since CCl 4 decreases rapidly with altitude, this difference is far more pronounced in relative terms.MIPAS shows slightly lower volume mixing ratios than ACE-FTS, in general.Part of this might be attributed to PAN not being accounted for in the ACE-FTS v3.5 retrieval (Harrison et al., 2017).With PAN missing from the forward model calculations, the retrieval increases CCl 4 to compensate.Preliminary ACE-FTS version 4 results indicate that retrieved CCl 4 will skew lower when PAN is included.However, Harrison et al. (2017) do not investigate the magnitude of the effect of including PAN versus not including it.Other items changed in the retrieval, e.g. the microwindow set and new cross sections, so it is not clear how much of the decrease in CCl 4 can be attributed to the inclusion of PAN as an interferer in the ACE-FTS retrieval.Nevertheless, the agreement between MIPAS Envisat and ACE is very good, staying within the 10 % range for the differences up to above 15 km.

MIPAS-B2
MIPAS-B2 is the follow-up of MIPAS-B (Friedl-Vallon et al., 2004), which was lost in 1992.MIPAS-B and MIPAS-B2 measurements add up to more than 20 flights to date.MIPAS-B2 covers the spectral range from 750 to 2500 cm −1 (13.3 and 4 µm) and vertical ranges up to the floating altitude of typically around 30-40 km.The vertical sampling is approximately 1.5 km.The spectral region used for the MIPAS-B2 retrieval ranges from 786.0 to 806.0 cm −1 .MIPAS-B2 Atmos.Meas. Tech., 10, 2727-2743, 2017 www.atmos-meas-tech.net/10/2727/2017/altitude range.The MIPAS-B2 measurement lies well within the spread of all collocated MIPAS Envisat profiles.The difference (middle panel) is always close to the total combined error, which includes all error estimates except the spectroscopy error.The latter has not been included because a MIPAS Envisat retrieval setup was used for this comparison which is based on the same spectroscopic data as the MIPAS-B2 retrieval.The right panel shows the relative error, which stays well within 5 % up to 17 km.Only between 16 and 18 km, the relative difference noticeably exceeds the combined error of the instruments.
The comparison of the MIPAS-B2 flight on 31 March 2011 (Fig. 15) with MIPAS Envisat presents even better agreement.The difference between the two profiles never exceeds 5 pptv (middle panel) and stays within or close to the combined error of the instruments throughout the whole altitude range.Larger deviations in the relative differences only occur above 18 km, where the combined error of the instruments also increases rapidly, because of small volume mixing ratios of CCl 4 .Overall, the comparisons with MIPAS-B2 show excellent agreement between the two instruments.This suggests that the MIPAS Envisat CCl 4 error estimates are realistic and that the residuals in the CO 2 lines mentioned in Sect.3.2 have no major impact on the CCl 4 retrieval.This is also supported by Fig. 10, at least up to about 18 km, since the standard deviation of the profiles can be explained by the MIPAS Envisat error estimates to a large extent.

Cryosampler
The cryosampler whose measurements are used here was developed at Forschungszentrum Jülich (Germany) in the early 1980s (Schmidt et al., 1987) and is a balloon-borne instrument.It collects whole air samples which are then frozen during the flight and analyzed using gas chromatography after the flight.In this analysis, a flight performed on 1 April 2011 by the University of Frankfurt (Fig. 16 black circles) is compared to collocated MIPAS Envisat profiles that lie within 1000 km and 24 h of the cryosampler profile.The MIPAS Envisat profiles used for the comparison are those retrieved with the new spectroscopic data set (continuous blue line: closest MIPAS profile; red line: MIPAS mean profile; bluegreyish lines: all collocated MIPAS profiles).In addition, the closest profile produced with the old spectroscopic data set is shown (dashed blue line).The only difference between the blue line and the dashed blue line are the different spectroscopic data sets.It is clearly visible that the closest MIPAS profile produced with the new spectroscopic data comes closer to the cryosampler measurements, even though these still show slightly lower volume mixing ratios of CCl 4 .A similar pattern of two outliers (second and forth lowest cryosampler measurements) was also seen in a comparison of cryosampler and MIPAS measurements of CFC-11 and CFC-12 (Eckert et al., 2016), even though the second lowest outlier is not as obvious for the CFCs.However, this might be an indication that cryosampler captured fine structures (like laminae) produced by the unique atmospheric situation in spring 2011 (Manney et al., 2011;Sinnhuber Atmos. Meas. Tech., 10, 2727-2743, 2017 www.atmos-meas-tech.net/10/2727/2017/

Figure 3 .
Figure 3. PAN altitude-latitude cross sections (July 2008) from a separate retrieval using the old CCl 4 distributions (a) and resulting from a joint retrieval with CCl 4 (b).

Figure 4 .
Figure 4. Impact of the CO 2 Q-branch at 11.5 km altitude without considering line mixing: (a) spectra; (b) residuals.Black: measured spectrum, hardly discernible because overplotted by modelled spectra.

Figure 5 .Figure 6 .
Figure 5. Impact of the CO 2 Q-branch at 11.5 km altitude taking line mixing it into account: (a) spectra; (b) residuals.Black: measured spectrum, hardly discernible because overplotted by modelled spectra.Note the different scale of the residual axis compared to Fig. 4.

Figure 8 .
Figure 8. Altitude-latitude cross sections of MIPAS CCl 4 for the RR period.(a-c) January 2010, March 2011 and April 2011.

Figure 11
Figure11shows an altitude-latitude cross section of MI-PAS Envisat CCl 4 trends.These trends were estimated by the same method as described byEckert et al. (2014), which is based on the method by von Clarmann et al. (2010).In addition to the setup used byEckert et al. (2014), the El-Niño-Southern Oscillation (ENSO) was also taken into account.The data set used for trend calculation covers the entire MIPAS Envisat measurement period from July 2002 to April 2012.The distribution of the trends agrees well with the trends estimated byValeri et al. (2017), who calculated trends from MIPAS Envisat V7 data they formerly retrieved and displayed them on a pressure-latitude grid.The most likely cause of differences between their and our trend es-

Figure 10 .
Figure10.Comparison of the estimated total error with the standard deviation of several MIPAS profiles for a quiescent atmospheric situation (Equator).Red: total error budget; blue: standard deviation.

Figure 11 .
Figure 11.Altitude-latitude cross sections of MIPAS CCl 4 trends covering the entire measurement period from July 2002 to April 2012.Red colours indicate increasing CCl 4 volume mixing ratios.Blue colours indicate declining CCl 4 concentrations.Hatching shows where no statistically significant trends could be calculated at 2σ confidence level.

Figure 13 .
Figure 13.Comparison of MIPAS Envisat and version 3.5 ACE-FTS CCl 4 .(a) Mean profiles of all coincident profiles (black: ACE-FTS; magenta: MIPAS).Dashed lines show the standard deviations of the mean profiles.(b) Number of coincident points per altitude.(c) Correlation coefficient of the mean profiles.(d) Relative differences of the mean profiles.(e) One standard deviation of the relative differences of the mean profiles.

Figure 14 .
Figure 14.Comparison of MIPAS Envisat and MIPAS-B2 CCl 4 for the MIPAS-B2 flight on 24 January 2010 over Kiruna, Sweden.(a) Mean profile of all coincident profiles (black line: MIPAS-B2; red line: MIPAS mean; red squares: coincident MIPAS measurements).(b) Absolute total error budget without consideration of the spectroscopy error.(c) Relative error budget − red continuous line: difference between the mean profiles; red dotted line: standard deviation; blue dotted line: mean combined precision; blue dashed line: total mean combined error.

Figure 15 .
Figure 15.Comparison of MIPAS Envisat and MIPAS-B2 CCl 4 for the MIPAS-B2 flight on 31 March 2011 over Kiruna, Sweden.(a) Mean profile of all coincident profiles (black line: MIPAS-B2; red line: MIPAS mean; red squares: coincident MIPAS measurements).(b) Absolute total error budget without consideration of the spectroscopy error.(c) Relative error budget − red continuous line: difference between the mean profiles; red dotted line: standard deviation; blue dotted line: mean combined precision; blue dashed line: total mean combined error.

Figure 16 .
Figure 16.Comparison of MIPAS Envisat and cryosampler CCl 4 .The cryosampler measurement was taken on 1 April 2011.The continuous and dashed blue lines are the respective closest MIPAS Envisat profiles calculated using the new and the old spectroscopic data set.

Table 2 .
Error estimates for a midlatitude profile during the FR period.Errors are given in pptv (relative errors in %).

Table 3 .
Error estimates for a midlatitude profile during the RR period.Errors are given in pptv (relative errors in %).
ATMOS CCl 4 mixing ratios also agree well withLiang et al. (2016, Fig. 2) where a time series of CCl 4 surface mixing ratios over several decades is shown.Volume mixing ratios at the lower end of the profiles are noticeably higher than 100 pptv, which is in very good agreement with peak values of CCl 4 shown inLiang et al. (2016,  Fig.

Table A1 .
Error estimates for an equatorial profile during the FR period.Errors are given in pptv (relative errors in %).

Table A2 .
Error estimates for a polar summer profile during the FR period.Errors are given in pptv (relative errors in %).

Table A3 .
Error estimates for a polar winter profile during the FR period.Errors are given in pptv (relative errors in %).

Table A5 .
Error estimates for a polar summer profile during the RR period.Errors are given in pptv (relative errors in %).

Table A6 .
Error estimates for a polar winter profile during the RR period.Errors are given in pptv (relative errors in %).