CCl 4 distribution derived from MIPAS ESA V 7 data : validation , trend and lifetime estimation

Atmospheric emissions of Carbon tetrachloride (CCl4) are regulated by the Montreal Protocol due to its role as a strong ozone-depleting substance. The molecule has been the subject of recent increased interest as a consequence of the so called “mystery of CCl4,” the discrepancy between atmospheric observations and reported production and consumption. Surface measurements of CCl4 atmospheric concentrations have declined at a rate almost three times smaller than its lifetimelimited rate, suggesting persistent atmospheric emissions despite the ban. In this paper, we study CCl4 vertical and zonal 5 distributions in the upper troposphere and lower stratosphere (including the photolytic loss region, 70-20 hPa), its trend, and its stratospheric lifetime using measurements from the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS), which operated onboard the ENVISAT satellite from 2002 to 2012. Specifically, we use the MIPAS data product generated with Version 7 of the Level 2 algorithm operated by the European Space Agency. The CCl4 zonal means show features typical of long-lived species of anthropogenic origin that are destroyed primarily 10 in the stratosphere, with larger quantities in the troposphere and a monotonic decrease with increasing altitude in the stratosphere. In the troposphere, the largest concentrations are observed at the latitudes of major industrial countries (20◦/50◦N). The good agreement we find between MIPAS CCl4 and independent measurements from other satellite and balloon-borne remote sounders proves the reliability of the MIPAS dataset. CCl4 trends are calculated as a function of both latitude and altitude. Negative trends are found at all latitudes in the upper15 troposphere / lower-stratosphere region, apart from a region in the Southern mid-latitudes between 50 and 10 hPa where the trend is positive. At the lowest altitudes sounded by MIPAS, we find trends consistent with those determined on the basis of long-term ground-based measurements. For higher altitudes, the trend shows a pronounced asymmetry between Northern and Southern Hemispheres, and the magnitude of the decline rate increases with altitude. At 50 hPa the decline is about 30-35%/decade, close to the lifetime-limited trend. 20 1 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-1163, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 2 March 2017 c © Author(s) 2017. CC-BY 3.0 License.


Introduction
Carbon tetrachloride (CCl 4 ) is a strong ozone-depleting substance with an ozone depleting potential of 0.72 and a strong greenhouse gas with a 100-year global warming potential of 1730 (Harris et al., 2014).Regulated by the Montreal Protocol, the production of CCl 4 for dispersive applications was banned for developed countries in 1996, while developing countries were allowed a delayed reduction with the complete elimination by 2010 (Liang et al., 2014).CCl 4 can still be legally used as a feedstock, for example in the production of hydro-fluorocarbons.
The dominant loss mechanism for atmospheric CCl 4 is through photolysis in the stratosphere.The other major sinks are degradation in the oceans and degradation in soil.The estimated partial lifetimes provided in the latest ozone assessment report (Carpenter et al., 2014) with respect to these three sinks are 44 years for the atmospheric sink, 94 years for the oceanic sink, and 195 years for the soil sink.The combination of these three partial loss rates yields a total lifetime estimate of 26 years.
CCl 4 atmospheric concentration is routinely monitored by global networks such as Advanced Global Atmospheric Gases Experiment (AGAGE, http://agage.mit.edu/)(Simmonds et al., 1998;Prinn et al., 2000Prinn et al., , 2016) ) and National Oceanic and Atmospheric Administration / Earth System Research Laboratory / Halocarbons & other Atmospheric Trace Species (NOAA / ESRL / HATS, http://www.esrl.noaa.gov/gmd/hats/).The concentration of CCl 4 has been decreasing in the atmosphere since the early 1990s, and the latest ozone assessment report (Carpenter et al., 2014) indicates that the global surface mean mole fraction of CCl 4 continued to decline from 2008 to 2012.AGAGE and University of California Irvine (UCI) networks report rates of decline of 1.2-1.3%yr −1 from 2011 to 2012, whereas the rate of decline reported by the NOAA/HATS network was 1.6% yr −1 .These relative declines in mole fractions at Earth's surface are comparable to declines in column abundances of 1.1-1.2%yr −1 (Brown et al., 2011;Rinsland et al., 2012).A significant discrepancy is observed between global emissions estimates of CCl 4 derived by reported production and feedstock usage (bottom-up emissions) compared to those derived by atmospheric observations (top-down emissions).This discrepancy has recently stimulated a particular interest in furthering the understanding of atmospheric CCl 4 .A study performed with a 3-D chemistry-climate model using the observed global trend and the observed inter-hemispheric gradient (1.5±0.2 ppt for 2000-2012) estimated a total lifetime of 35 years (Liang et al., 2014).Recently, a study has reassessed the partial lifetime with respect to the soil sink to be 375 years (Rhew and Happell, 2016), and another study has reassessed the partial lifetime with respect to the ocean sink to be 209 years (Butler et al., 2016).These new estimates of the partial lifetimes with respect to soil and oceanic sinks produce a new total lifetime estimate of 33 years, consistent with the estimate given in Liang et al. (2014).This longer total lifetime reduces the discrepancy between the bottom-up and top-down emissions from 54 Gg yr −1 to 15 Gg yr −1 (SPARC, 2016).While the new bottom-up emission is still less than the top-down emission, the new estimates reconcile the CCl 4 budget discrepancy when considered at the edges of their uncertainties.A recent study estimated that the average European emissions for 2006-2014 were 2.3 Gg yr −1 (Graziosi et al., 2016), with an average decreasing trend of 7.3% per year.
Since the atmospheric loss of CCl 4 is mainly due to photolysis in the stratosphere, satellite measurements that provide vertical profiles are particularly useful in validating the stratospheric loss rates in atmospheric models.A global distribution of CCl 4 extending up to the mid-stratosphere was obtained by the Atmospheric Chemistry Experiment-Fourier Transform Spectrometer (ACE-FTS) (Allen et al., 2009).This study derived an atmospheric lifetime of 34 years through correlation with CFC-11.Another determination of the atmospheric lifetime of CCl 4 was produced with ACE-FTS measurements in Brown et al. (2011), where the lifetime is estimated to be 35 years.A trend of atmospheric CCl 4 from ACE-FTS measurements was reported in Brown et al. (2013), averaged in the 30 • S/30 • N latitude belt and in the altitude range from 5 to 17 km, were it was found to be decreasing at a rate of 1.2% yr −1 .
In this paper, we report the global atmospheric distribution of CCl 4 as a function of altitude and latitude obtained from the measurements of the limb emission sounder MIPAS (Michelson Interferometer for Passive Atmospheric Sounding) (Fischer et al., 2008) onboard the ENVISAT satellite.The data product employed here was generated with the processor of the European Space Agency (ESA) Version 7 (ESA, 2016).MIPAS CCl 4 vertical profiles are validated by comparing them with correlative independent measurements.The trend of CCl 4 as a function of altitude and latitude is also determined.The MIPAS measurements provide a denser and more complete geographical coverage than those provided by the ACE-FTS measurements, allowing a more precise knowledge of the CCl 4 global distribution and of the trend.The key photolytic loss region (70-20 hPa) is also analyzed.
In Section 2, we introduce MIPAS measurements, the retrieval setup, and the error budget of the CCl 4 profiles.In Section 3, we discuss the global CCl 4 distribution and the inter-hemispheric differences determined from MIPAS measurements.In Section 4, we show the results of the comparisons between MIPAS and CCl 4 correlative measurements from the balloon version of the MIPAS instrument and the ACE-FTS.In Section 5, we illustrate the method adopted for the estimation of the atmospheric trends and the results of trend analysis, along with some comparisons to previously published results.In Section 6, we evaluate the CCl 4 stratospheric lifetime using the tracer-tracer linear correlation method and compare the results with previously published estimates.

MIPAS measurements
In the first two years of operations (from July 2002 to March 2004) MIPAS acquired, nearly continuously, measurements at Full spectral Resolution (FR), with a spectral sampling of 0.025 cm −1 .On 26 March 2004, FR measurements were interrupted due to an anomaly in the movement of the interferometer drive unit.After instrument diagnosis and tests by the hardware experts, atmospheric measurements were resumed in January 2005.After this date, however, MIPAS adopted a reduced spectral resolution of 0.0625 cm −1 .Being achievable with a shorter interferometric scan, measurements with this spectral resolution require a reduced measurement time compared to the FR, thus allowing a finer spatial sampling.For this reason, the measurements acquired from January 2005 onward are referred to as Optimized Resolution (OR) measurements.Compared to the FR, they show both a reduced Noise Equivalent Spectral Radiance (NESR), and finer vertical and horizontal spatial samplings.
The nominal FR (OR) scan pattern consists of 17 (27) sweeps with tangent heights in the range from 6-68 (7-72) km with 3 (1.5)km steps in the Upper Troposphere / Lower Stratosphere (UTLS) region.Full details of the MIPAS measurements acquired in the two mission phases are reported in Raspollini et al. (2013).It is worth mentioning here that in both mission phases MIPAS measurements cover the whole globe with a dense sampling, facilitating detailed studies on the evolution of atmospheric composition.The ESA operational Level 2 algorithm retrieves target parameters at the tangent points of the limb measurements (or at a subset of them).The inversion process minimizes the χ 2 -function, using the Gauss-Newton iterative scheme with the Marquardt modification.An adaptive a-posteriori regularization is used in order to smooth the profiles with a strength determined on the basis of the error bars of the unregularized profile (Ceccherini, 2005;Ceccherini et al., 2007;Ridolfi andSgheri, 2009, 2011).The ESA Level 2 processor version 7 retrieves CCl 4 volume mixing ratio (VMR) profiles simultaneously with a set of other target parameters.The retrieval is based on the fit of a set of narrow (3 cm −1 ) spectral intervals called microwindows (MWs) containing relevant information on the target parameters.As for all MIPAS ESA retrievals, the MWs for CCl 4 retrievals are selected with the MWMAKE algorithm (Dudhia et al., 2002).This algorithm identifies the spectral intervals to be used in the inversion, with the aim of minimizing the total retrieval errors (including both systematic and random components).The MWs used in the ESA Level 2 retrievals from nominal FR and OR measurements are listed in Table 1.
CCl 4 VMR is retrieved only up to about 27 km, since above this altitude the CCl 4 concentration is too small to generate a sufficient contribution to the measured spectrum for analysis.Furthermore, to avoid numerical instabilities in the inversion of OR measurements (that vertically oversample the limb), the retrieval grid includes only one out of every two tangent points.
Fig. 1 characterizes a typical CCl 4 retrieval from nominal limb scans acquired in the FR (top panel) and OR (bottom panel) measurement phases.The coloured solid lines show the 8 rows (corresponding to the 8 retrieval grid points) of the Averaging Kernels (AKs).Typically the number of degrees of freedom of the retrieval (trace of the AK matrix) is 5-6 for FR and 4-5 for OR measurements.The slightly smaller number of degrees of freedom obtained in the OR retrievals stems from the fact that, to make the retrieval more stable, CCl 4 is not retrieved at every tangent point of the OR limb measurements.The dotted red line of Fig. 1 represents the vertical resolution, calculated as the Full Width Half Maximum (FWHM) of the AK rows.

Error budget
To evaluate the CCl 4 VMR error due to the mapping of the measurement noise in the retrieval we use the error covariance matrix provided by the retrieval algorithm (Ceccherini and Ridolfi, 2010).The other error components affecting the individual CCl 4 VMR profiles are evaluated at Oxford University using the MWMAKE tool.Fig. 2 summarizes the most relevant error components affecting each individual retrieved CCl 4 profile, using the MWs of Table 1, for both the FR (top panel) and OR (bottom panel) nominal MIPAS measurement cases.
The key "RND" in the plots refers to the mapping of the measurement noise in the retrieval, as evaluated for typical FR and OR retrievals.Apart from the "NLGAIN" error that will be discussed later, the other error components, in both the FR and OR cases, can be grouped as follows: a) the errors due to the uncertainties in the (previously retrieved) pressure and temperature The main errors of type a) are due to interfering gases whose VMRs are retrieved before CCl 4 with some random error.
Therefore, like the RND error component, they change randomly from profile to profile.Thus, in the calculated (monthly) averages they scale down with the inverse square root of the number of averaged profiles.in Castelli et al. (2016), cause systematic (and opposite in sign) differences between profiles retrieved from measurements acquired in the ascending and the descending parts of the satellite orbits.These errors cancel-out when calculating averages that evenly include profiles retrieved from measurements belonging to the ascending and the descending parts of the orbits.Errors of type c) are constant and may cause profile biases, however, being constant, they do not affect the calculated trends.Regarding the errors due to the imperfect instrument line-shape modeling (type d), since the gain of MIPAS bolometric detectors remained constant throughout the whole mission, there is no hint of a possible degradation of instrument optics and thus of a possible change in the instrument line-shape.This type of error, therefore, also has no impact in trend calculations.
Imperfect instrument radiometric calibration also causes an error.This error is plotted in Fig. 2 with the label "NLGAIN".
Being of the order of 0.4% in the upper part of the retrieval range, it is rather small in individual CCl 4 profiles.Although small, this error is important when calculating atmospheric trends as it includes the uncertainty in the correction applied to the radiances to account for the non-linearities of MIPAS photometric detectors (Kleinert et al., 2007).In MIPAS Level 1b radiances up to version 5, the applied non-linearity correction is constant throughout the whole MIPAS mission.However, non-linearities change over the course of the mission due to progressive ageing of the detectors.A constant correction implies, therefore, a drift of the radiometric calibration error during the mission, with a direct impact in the calculated trends.MIPAS Level 1b radiances version 7 overcome this problem as they use a time-dependent non-linearity correction scheme.The residual drift of the calibration error after this time-dependent correction is still being characterized; however, preliminary results (Birk priv. com. 2016) show that it is smaller than 1% across the entire mission.MIPAS Level 1b radiances version 5 were used in the past to extract information on trends of different gases, either ignoring this effect (see, e.g., CFC-11/CFC-12 in Kellmann et al. (2012), or HCFC-22 in Chirkov et al. (2016)) or correcting the drift via intercomparison with other instruments assumed to be drift-free (Eckert et al., 2014).Recently it has been shown (Eckert et al., 2016) that ignoring this effect introduces a significant error on the trend estimation.The MIPAS Level 1b calibrated radiances version 7 employed here are considered a significant improvement from the point of view of the correction of this drift.The general good quality of fits obtained in CCl 4 retrievals is illustrated in Fig. 3.The figure refers to the MWs used in the FR retrievals.We are not showing the residuals in the single MW used for OR retrievals as it mostly overlaps the third MW of FR retrievals.The upper plot of Fig. 3 shows the average of 1141 observed (black dots) and simulated (red line) limb radiances in the MWs used for CCl 4 retrievals.The averages include spectra with tangent heights in the range from 6 to 17 km.The lower plot shows the average residuals of the fit (observation minus simulation, blue line) as well as the average noise level 5 of the individual MIPAS measurements (dashed lines).The grey areas indicate spectral channels that, as recommended by the MWMAKE algorithm, are excluded from the fit to minimize the total retrieval error.Note that the average residuals shown in Fig. 3 have an associated random error given by the noise of the individual measured spectra divided by the square root of the number of averaged spectra, i.e. ≈ 1nW/(cm 2 sr cm −1 ).This implies that while the amplitude of the average residuals is not FR retrievals.The averages include spectra with tangent height from 6 to 17 km.The lower plot shows the average residuals of the fit (blue line, observation minus simulation) as well as the average noise level of the individual measurements (dashed lines).The grey areas indicate spectral channels excluded from the fit.The radiance units (r.u.) in the vertical axes of the plots are nW/(cm 2 sr cm −1 ).

CCl 4 global distribution
Figure 4 shows the global monthly distribution of MIPAS CCl 4 VMR for a representative month from each of the four seasons, spanning the time period from August 2010 through May 2011.Here, retrieved profiles were first interpolated to fixed pressure 5 levels (see Sect.5.1), and then binned in 5 • latitude intervals.In all the considered months, the zonal averages show the typical shape of the long-lived species of anthropogenic origin, which are emitted at the surface and destroyed primarily in the stratosphere.Larger values are found in the troposphere, and then the VMR monotonically decreases with increasing altitude in the stratosphere.In the lower stratosphere, concentrations between 30 • S and 30 • N are significantly larger compared to those at higher latitudes.This pattern can be attributed to the Brewer-Dobson circulation that is responsible for the uplift of the surface spring begins in March in the North Pole).This was previously observed for other long lived anthropogenic species (Kellmann et al., 2012).The effect is larger in the Antarctic.The maps in Fig. 5 show the time evolution of CCl 4 at all latitudes from July 2002 to April 2012.The three maps refer to different pressure levels: 50 hPa (upper map), 90 hPa (middle map) and 130 hPa (lower map).The seasonal variability is also clearly visible from the maps of Fig. 4, with opposite phases in the two Hemispheres, more pronounced at mid-latitudes and in the polar regions.As previously mentioned, in the troposphere a minimum is found in the Northern and Southern Polar Spring.
Modulated by this seasonal variability, at all altitudes a constant trend and an inter-hemispherical gradient can also be observed and are further analysed in the subsequent figures.We also notice that for pressures larger than 100 hPa, the CCl 4 measured in the OR phase has a positive bias with respect to that measured in the FR phase.This bias, discussed also in Sect.4.1, may be due to the different MWs used for the retrieval in the two mission phases, or to the different limb sampling patterns adopted.emissions (Lovelock et al., 1973;Liang et al., 2014).Anthropogenic emissions are larger in the NH, and the transport of these hemispheric differences.At higher altitudes, the asymmetry between the North and South CCl 4 VMR depends not only on the unbalanced emissions at the surface but also on the general circulation of the atmosphere and on seasonal transport.Indeed, we have seen that, especially at high latitudes, N-S differences strongly depend on the season.The large differences visible in Fig. 6 at high latitudes in the stratosphere originate from the subsidence effect during polar winter and spring, bringing  2004)) and by the ACE-FTS onboard the SciSat-1 satellite (Bernath et al., 2005).In the next two sub-sections we compare MIPAS CCl 4 profiles with co-located profiles obtained from these two instruments.

Comparison with MIPAS balloon
Stratospheric balloon measurements are particularly suitable for the validation of space-borne limb sounding instruments since these instruments are able to sound the atmosphere with high vertical resolution.The balloon-borne limb emission sounder MIPAS-B can be regarded as a precursor of the MIPAS satellite instrument (Friedl-Vallon et al. (2004) and references therein).
However, for other parameters the MIPAS-B performance is superior, in particular for the NESR and for the line of sight stabilization, which is based on an inertial navigation system supplemented with an additional star reference system and leads to a knowledge of the tangent altitude on the order of 90 m (3σ Retrieval of all species is performed on a 1 km grid with a least squares fitting algorithm using analytical derivative spectra calculated by the Karlsruhe Optimized and Precise Radiative transfer Algorithm (Höpfner et al., 2002;Stiller et al., 2002).
To avoid retrieval instabilities due to oversampling of vertical grid points, a regularization approach according to the method described by Tikhonov and Phillips constraining with respect to a first derivative a priori profile is adopted.The spectral window used for the MIPAS-B target parameter retrieval of CCl 4 covers the 786.0-806.0cm −1 interval.Spectroscopic parameters for the calculation of the infrared emission spectra are a combination of the HITRAN 2008 (Rothman et al., 2009) database and the MIPAS dedicated database (Raspollini et al., 2013;Perrin et al., 2016).The CCl 4 cross sections are taken from HITRAN as in MIPAS/ESA retrievals version 7. The MIPAS-B error budget includes random noise as well as covariance effects of the fitted parameters, temperature errors, pointing inaccuracies, errors of non-simultaneously fitted interfering species, and spectroscopic data errors (1σ).For CCl 4 the precision error is estimated to be between 5-10%, while the total error is 11-15%.Further details on the MIPAS-B data analysis and error estimation are provided in Wetzel et al. (2012) and reference therein.Table 2 lists all the MIPAS-B flights used for intercomparison with MIPAS on ENVISAT.
Further than the direct matches where the balloon and the satellite instruments observe simultaneously (within pre-defined margins) the same air-masses, we also considered trajectory matches.In this case both forward and backward trajectories were calculated (Naujokat and Grunow, 2003) by the Free University of Berlin from the balloon measurement geolocation to search for air-masses sounded by the satellite instrument.Temperature and VMR values from the satellite profiles were interpolated to the trajectory match altitude such that these values can be directly compared to the MIPAS-B data at the trajectory start point altitude.To identify both direct and trajectory matches, a coincidence criterion of 1 hour and 500 km was adopted.PAS/ENVISAT mission, respectively.Combined random, systematic and total errors are also shown in the plots.The numbers reported on the left side of the plots indicate the number of matching profiles contributing to the statistics.The results of the intercomparison can be summarized as follows.In the case of FR measurements: for pressures between 80 and 190 hPa MI-PAS/ENVISAT shows a statistically significant negative bias of about −10% with respect to MIPAS-B, this bias is however within the combined total error bounds.A statistically significant positive bias is also evident for pressures smaller than 25 hPa.

5
It increases with altitude and quickly becomes incompatible with the total combined error.This bias, however, is not a major concern because it is localized at the upper end of the retrieval range.In this region the predicted uncertainty is so large that the linear approximation of the error propagation theory may easily fail to explain the discrepancies between the measurements of the two instruments.In case of OR measurements: for pressures between 150 and 190 hPa MIPAS/ENVISAT shows a statistically significant positive bias of about +10% with respect to MIPAS-B; this bias is however within the combined total and, for pressures smaller than 20 hPa is no longer compatible with the total combined error.As in the FR case, this large bias occurs at the upper end of the MIPAS/ENVISAT retrieval range where the predicted combined error is very large.Furthermore, validation with ACE (see next Section) indicates a negative bias of MIPAS wrt ACE-FTS, in the same altitude region, hence MIPAS/ENVISAT is in the middle between MIPAS balloon and ACE-FTS.

Comparison with ACE-FTS V3.5
5 ACE-FTS is a Canadian solar occultation limb sounder operating since 2004 from SciSat in a low (≈ 650 km) circular orbit.The measured spectra cover the region from 750 to 4400 cm −1 with a spectral resolution of 0.02 cm −1 (Bernath et al., 2005).Several target atmospheric parameters are routinely retrieved from ACE-FTS measurements.Among them, temperature, pressure, and the VMR profiles of over 30 atmospheric trace gases and over 20 subsidiary isotopologues.Profiles are retrieved in the range from ∼ 5 to 150 km, with a vertical field of view of ∼ 3-4 km and a vertical sampling of 2-6 km.The ACE-FTS retrieval Several hundred ACE-FTS measurements are coincident with MIPAS soundings of the OR part of the mission.These measurements are located both in the Northern and Southern hemispheres, mainly at latitudes larger than 45 • .For comparison with MIPAS, all ACE-FTS CCl 4 data used were screened using the v3.5 quality flags.As recommended by Sheese et al. (2015), any profile data point with flag value of 2 or greater was removed and any profile containing a flag value between 4 and 7, inclusive, was discarded.For intercomparison with MIPAS measurements we adopted a matching criterion of 3 hours and 300 km.We also tested different matching criteria, such as 2 hours and 300 km, 3 hours and 200 km, but found no significant changes in the validation results.First we interpolated the matching MIPAS and ACE-FTS CCl 4 profiles to a fixed set of pressure levels.Then we grouped the profile differences in latitudinal intervals.The results of the comparison are summarized in Fig. 9.Each of the four plots of the figure refers to one of the considered latitude intervals: 50-70 • and 70-90 • in both the Southern and the Northern hemispheres.Each plot shows the average CCl 4 difference profile between co-located MIPAS and ACE-FTS measurements (red) with standard deviation of the mean (red error bars).The standard deviation of the difference (orange), the total random error (green), the total systematic error of the difference (blue) are also shown.The number of co-located pairs contributing at each pressure level is reported on the right side of each plot.The average difference (red line) quantifies the systematic bias between ACE-FTS and MIPAS, the error bars indicate its statistical significance.The standard deviation (orange) is an ex-post estimate of the combined random error of the individual profile differences and, therefore, should be similar to its ex-ante estimate represented in the plots by the green line.We calculated the ex-ante random error of the individual profile differences as the quadrature summation of the ACE-FTS and MIPAS random errors.The ACE-FTS random error is estimated via the noise error covariance matrix of the retrieval included in the Level 2 products.The MIPAS random error is estimated as the quadrature summation of the measurement noise error evaluated by the covariance matrix of the retrieval (Ceccherini and Ridolfi, 2010) and the other error components that are expected to change randomly in our sample, i.e. the errors that we classified of types a) and b) in Sect.2.1.The systematic error of the profile differences is obtained as the quadrature summation of the ACE-FTS and the MIPAS errors that are constant within the sample and are not expected to bias in the same direction the measurements of the two instruments.On the basis of the error figures suggested by Allen et al. ( 2009), for ACE-FTS we assumed a 20% systematic error constant at all pressure levels.For MIPAS we calculated the quadrature summation of systematic errors that in Sect.2.1 we classified as of type c) and d).For the calculation of the combined systematic error we explicitly excluded the uncertainty in the CCl 4 cross-section data (Rothman et al., 2005) that are used, approximately in the same spectral region, both in MIPAS and ACE-FTS retrievals.
Apart from the latitude interval from 50 to 70 • S, the systematic differences between MIPAS and ACE-FTS are within 5 pptv (∼ 10 %, mostly not significant from the statistical point of view) in the pressure range from 50 to 100-110 hPa.The amplitude of systematic differences increases up to 15-20 pptv and becomes statistically significant at 30 hPa, while it is again quite small at 20 hPa.In the latitude interval from 50 to 70 • S we observe a statistically significant ≈ 10 pptv low bias of MIPAS with respect to ACE-FTS, almost uniform over the entire retrieval height range.At all latitudes, the observed biases are compatible with the estimated combined systematic error only for pressures greater than 40 hPa.At 30 hPa the bias is statistically significant and incompatible with error bars.The reason for this inconsistency is still unclear; however, preliminary 5 investigations show that the inconsistency will be reduced when using the future release version 4.0 of ACE-FTS products.
The ex-ante estimate of the combined random error (green line in Fig. 9) agrees pretty well with the ex-post estimated standard deviation of the profile differences (orange line) in the range between 40 and 80-100 hPa.At the edges of the retrieval range the observed variability of the differences generally exceeds the ex-ante estimate of the random error.This may be due 16 Atmos.Chem. Phys. Discuss., doi:10.5194/acp-2016-1163, 2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2017 c Author(s) 2017.CC-BY 3.0 License.both to the fact that our ex-ante random error estimate does not take into account the imperfect matching of the compared profiles, and to the fact that, at these specific altitudes, the sensitivity of the measurements to the CCl 4 VMR is so low that the linear approximation of the error propagation theory could provide only rough error estimates.
As a final remark we note that at 30 hPa MIPAS-B (Fig. 8) and ACE-FTS (Fig. 9) intercomparisons provide contrasting indications on the MIPAS bias in the OR part of the mission.While MIPAS-B suggests a positive MIPAS bias of about 10 pptv, ACE-FTS points to a negative bias of 10 − 20 pptv.

Trend calculation method
The measurements used for the analysis presented in this study cover the entire MIPAS mission, from July 2002 to April 2012.
The CCl 4 VMR profiles considered are those derived by the ESA Level 2 processor version 7 analysing MIPAS limb scanning measurements with tangent heights in the 6-70 km range, obtained from nominal (NOM), middle atmosphere (MA) and Upper Troposphere Lower Stratosphere (UTLS1) observational modes (Raspollini et al., 2013).
First we linearly interpolate in log-pressure all the considered CCl 4 VMR profiles to the 28 SPARC (Stratospheric Processes and their Role in Climate) data initiative (Hegglin and Tegtmeier, 2011) pressure levels (300, 250, 200, 170, 150, 130, 115, 100, 90, 80, 70, 50, 30, 20, 15, 10, 7, 5, 3, 2, 1.5, 1.0, 0.7, 0.5, 0.3, 0.2, 0.15, 0.1 hPa).We then group the interpolated profiles in 5 • latitude bins and calculate monthly averages.Finally, using the least-squares method, for each latitude bin and pressure level we fit the following function VMR(t) to the time series of the monthly averages: In this expression t is the time expressed in months since the beginning of the mission (July 2002) and a FR , a OR , b, f 1 , f 2 , g and c i , d i , i = 1, ..., 8 are the 22 fitting parameters.The function 1 P (t) is the indicator function of the time interval P, such that 1 P (t) = 1 if t ∈ P and 1 P (t) = 0 otherwise.The functions qbo30(t) and qbo50(t) are the quasi-biennial oscillation (QBO) quantifiers and SRF(t) is the solar radio flux index.The two QBO terms (available at http://www.geo.fu-berlin.de/met/ag/strat/produkte/qbo/index.html)represent the Singapore winds at 30 and 50 hPa (Kyrölä et al., 2010).The SRF index is calculated using measurements of the solar flux at 10.7 cm (available at http://lasp.colorado.edu/lisird/tss/noaa_radio_flux.html) and is considered a good proxy for the solar activity.We re-normalized both the QBO and the SRF proxies to the interval [−1, +1] within the time frame covered by MIPAS mission.The terms in the sum are 8 sine and 8 cosine functions.They represent periodic oscillations with period T i .In T i we include annual (12 months), semi-annual (6 months) and other characteristic atmospheric periodicities of 3, 4, 8, 9, 18 and 24 months.We decided to fit two different offset parameters for the two parts of the mission: a FR for the FR and a OR for the OR part.The aim of this choice is to account for possible relative biases between the two phases of the mission.These may be caused, for example, by the different spectral resolutions adopted, by the different MWs used for the retrieval and by the different vertical and horizontal samplings of the instrument in the two mission phases.We calculate the uncertainty on the fitted parameters assuming each monthly average is affected by an error given by the standard deviation of the mean.Furthermore we multiply the uncertainty obtained from the error propagation analysis by the square root of the normalized least squares (the so-called "reduced χ 2 ").This latter operation is intended to account also for the quality of the fit in the evaluation of trend errors.

Results
Figure 10 shows some examples of CCl 4 trend analysis.Each panel refers to a specific latitude band and pressure level.The top plot of each panel shows the time series of the monthly averages with error bars given by the standard deviation of the mean (blue symbols).The red curve represents the best fitting function VMR(t), while the green line represents the constant and the linear (trend) terms of VMR(t).In the lower plot of each panel we show the residuals of the fit (the monthly averages minus the values calculated on the fitting curve).In each panel we also report the value obtained for the trend, its uncertainty and the difference between the two offset terms a F R − a OR .
The quality of the fit is generally better in the OR period.Indeed, in this mission phase the instrument provides measurements with more uniform and finer geographical coverage.We also carried-out a careful spectral analysis of the residuals of the fits.
Although not reported here for reasons of brevity, this analysis reveals that all the periodicities embedded in the considered time series of monthly means are properly accounted for by the fitting function (1).
Figure 11 summarizes the results obtained for CCl 4 trends.Panel a) shows the absolute trends.Negative trends are observed at all latitudes in the UTLS region.The magnitude of the negative trend decreases with increasing altitude.The trend shows slightly positive values (about 5-10 pptv/decade) in a limited region, particularly in the Southern mid-latitudes between 50 and 10 hPa.This feature is probably related to the asymmetry in the general circulation of the atmosphere.The air at higher altitudes can be considered older than the tropospheric air that has been lifted up by strong convection mechanisms in the tropical regions (Stiller et al., 2012).The tropospheric air just injected into the stratosphere is richer in CCl 4 .We attribute positive stratospheric trend values in certain latitude regions to the less effective mixing mechanisms in the stratosphere as compared to the troposphere at these latitudes.Similar features have also been observed by other authors in CFC-11 and CFC-12 trends (Kellmann et al., 2012).
Assuming for each latitude bin and pressure level the average CCl 4 VMR obtained from the full MIPAS dataset, we also calculated the relative CCl 4 trends.They are shown in the panel b) of Fig. 11.The same considerations made for the absolute trends apply also to relative trends.The asymmetry between the NH and the SH is very pronounced, the NH having larger negative relative trends increasing with altitude and reaching 30-35%/decade at 50 hPa.Note however that above 50 hPa they show large variations with both latitude and pressure.These oscillations correspond to extremely small average VMR values that make the relative trend numerically unstable.Panels c) and d) of Fig. 11 show, respectively, the absolute and percentage random errors on the trends.The uncertainties increase above 20 hPa.Large uncertainties are associated to latitude bins and pressure levels for which a relatively small number of measurements is available.For clarity in Fig. 12 we show the ratio  between CCl 4 trends and the related random errors.Latitude bins / pressure levels with ratio values less than 2 are marked with white and grey colors and correspond to trend values that are not significantly different from zero from the statistical point of view.Note, however, that most of the calculated trends are greater than 5 times the related error, and are thus statistically significant.In the maps of Fig.s 11 and 12, values corresponding to errors greater than 30% are masked with dashes.We consider unreliable any trends reported here with errors greater than this threshold.

Comparison with CCl 4 trends reported in literature
Although measurements acquired at ground stations cannot be directly compared with MIPAS profiles that have a lower altitude limit of 5-6 km, we can still compare tropospheric CCl 4 trends derived from MIPAS with trends derived from groundbased measurements.In particular, we consider observations provided by two networks that regularly perform long-term, highly accurate near-surface measurements of various tracers, including CCl 4 : the NOAA/ESRL/HATS (http://www.esrl.noaa.gov/gmd/hats/) and the AGAGE (Simmonds et al., 1998;Prinn et al., 2000Prinn et al., , 2016) ) http://agage.mit.edu/)networks.The NOAA/ESRL/HATS group provides accurate measurements of CCl 4 through three different programs: two in situ electron capture detector (ECD) measurement programs and one flask system using gas chromatography with ECD program.In this work we use a CCl 4 combined dataset, developed by the NOAA to homogenize all of the measurements made by the different programs (more details at http://www.esrl.noaa.gov/gmd/hats/combined/CCl4.html).All the CCl 4 NOAA records are reported on the NOAA-2008 scale.AGAGE measurements used here are obtained using in situ gas chromatography with ECD and reported on the SIO-2005 calibration scale.NOAA and AGAGE in situ measurements at common sites are inter-compared every 6 months for validation purposes.
To compare MIPAS CCl 4 trends to those derived from the ground-based measurements of NOAA and AGAGE, we first choose a pressure level belonging to the troposphere, with the following procedure.For each latitude bin (λ) and MIPAS monthly average profile we identify the tropopause with the pressure level where the monthly average temperature shows its minimum value.We multiply this pressure by 1.6 and find the nearest pressure level (p t (λ)) in the fixed pressure grid defined in Sect.5.1.For each latitude bin and month we then compute the monthly CCl 4 average at p t (λ).Finally, for each latitude bin, we calculate the trend at this month-and latitude-dependent tropospheric pressure as explained in Sect.In Table 3 we compare MIPAS tropospheric CCl 4 trends with trends derived for the 2002-2012 decade from NOAA/AGAGE stations located in the same latitude band.As we can see, some stations produce CCl 4 trends in very good agreement with MIPAS.However, in general, and especially in the polar regions, the variability of the tropopause is quite large, thus producing time series of MIPAS monthly averages at p t (λ) that can not be adequately matched by the fitting function defined in Eq. 1.
This feature sometimes generates large residuals in the trend fit and thus large trend errors and/or unrealistic trend values.

5
Despite this difficulty, from the statistical point of view the only trends calculated at the CGO site disagree significantly.We attribute this disagreement to the instabilities occurring in MIPAS data at low altitudes.Indeed, the MIPAS tropospheric trend estimated for the latitude bin 35 • /40 • S (the bin adjacent to the CGO site) is already equal to −9.16 ± 2.03 pptv/decade, i.e. in perfect agreement with the trend calculated from the CGO measurements.
Looking at the literature we found that Brown et al. (2011)  yearly averages of CCl 4 VMR in the altitude range from 5 to 17 km and fit the seven 2004-2010 yearly averages with a linear least-squares approach.The resulting trend is −13.2 ± 0.9 pptv/decade.If we average MIPAS trends presented in Sect.5.2 in the 30 • S/30 • N latitude interval and in the 100-300 hPa pressure range, with a filter discarding trend values with relative error greater than 30%, we get an average trend of −12.80 ± 0.12 pptv/decade.This value is in very good agreement with the trend determined from ACE-FTS.Note also that, since MIPAS measures atmospheric emission its sampling is finer than that of ACE-FTS both in space and time.With MIPAS it is therefore possible to achieve a smaller trend error.

Lifetime
In this Section we estimate the stratospheric lifetime of CCl 4 using tracer-tracer linear correlations in the lower stratosphere, as described in Plumb and Ko (1992).The lifetime can be calculated using the following equation: where τ 1 and τ 2 are the stratospheric lifetimes of the two selected tracers, dσ 2 /dσ 1 is the slope of the correlation and σ 1 , σ 2 are the tropospheric VMRs of the two species.The tracers with known lifetime employed in this study are CFC-11 and CFC-12.As suggested in Brown et al. (2013), for this study we consider only the latitudes in the so-called surf zone (Volk et al., 1997), between 30 • N/S and 70 • N/S.The tropical regions are not suitable to estimate the stratospheric lifetime using the tracer-tracer method due to the intense large-scale upwelling (Plumb and Ko, 1992).Similarly, the polar regions are not suitable for this study due to the intense subsidence, especially during winter (Plumb, 2007).For each month of the MIPAS mission and each 5 • latitudinal band between 30 • N/S and 70 • N/S, we determine the pressure level corresponding to the tropopause, taken as the level with a minimum in the monthly average temperature profile.For each latitudinal band, we determine the slope of the correlation considering only the VMR monthly averages of  and CCl 4 at the subset of SPARC pressure levels (see Sect. 5.1) above the tropopause.As suggested in Brown et al. (2013), we apply additional filters to the VMRs of CFC-11 and CFC-12.We consider only VMRs greater than 100 pptv for CFC-11 and greater than 300 pptv for CFC-12.This approach makes us confident that the calculated slope is not affected by VMR values associated with the upper stratosphere.We compute the tropospheric VMR values of the three species by averaging the MIPAS retrieved VMR values at the SPARC pressure level below the tropopause.For CFC-11 we assume a lifetime of 52(43 -67) years and for CFC-12 a lifetime of 102(88 -122) years, as recommended in SPARC (2013).The results of our analysis are reported in Table 4.The global average lifetime turns out to be 46(38 -60) years considering CFC-11 as the correlated tracer, and 48(41 -58) years considering CFC-12.These estimates are consistent with the most recent literature that suggests an atmospheric lifetime of 44(36 -58) years (SPARC, 2013(SPARC, , 2016)).Several older studies report atmospheric CCl 4 lifetimes between 30 and 50 years (Singh et al., 1976;Simmonds et al., 1988;Montzka et al., 1999;World Meteorological Organization (WMO), 1999;Allen et al., 2009).In Brown et al. (2013)  years in the SH) but they are not able to provide a solid justification for this finding.Again, the differences with our results are partially explained with the different CFC-11 lifetime considered (using the same CFC-11 lifetime (World Meteorological Organization (WMO), 2011) we would obtain a CCl 4 lifetime of 41±8 years in the NH and 39±6 years in the SH) but the choice of different reference lifetimes does not explain the hemispheric asymmetry reported in Brown et al. (2013).We also calculated CCl 4 lifetimes in the two hemispheres considering CFC-12 as correlated tracer and we obtained a lifetime of 47(41 -57) years in the NH and of 49(42 -59) years in the SH.

Conclusions
The ESA Version 7 processor has been used to retrieve for the first time the CCl 4 VMR global distribution in the UTLS using MIPAS measurements.The CCl 4 MIPAS observations cover the altitude range from 6 to 27 km and, having been obtained from emission measurements, provide a global coverage.The zonal means of CCl 4 VMR show features typical of long-lived species of anthropogenic origin that are destroyed primarily in the stratosphere by photolysis.The highest VMR values are found in the troposphere, and VMR monotonically decreases with increasing altitude in the stratosphere.In the lower stratosphere, the largest values are observed between 30 • S and 30 • N due to the intense updraft that occurs in the tropical region.The CCl 4 global distribution shows also a seasonal variability.This seasonality is more evident in the polar regions due to CCl 4 -poor mesospheric air subsidence induced by the polar vortex.
We calculated the IHG in the UTLS using average CCl 4 profiles measured in the NH and in the SH.At high latitudes in the stratosphere, differences the order of 10-15 pptv are observed, showing a strong seasonal variation, caused by subsidence (generally more intense in the SH) during polar winter and spring.At mid-latitudes, the CCl 4 VMR differences between the two hemispheres are less dependent on the season and are mainly caused by the larger CCl 4 emissions in the NH.

Figure 2 .
Figure 2. Main error components of the individual retrieved CCl4 VMR profiles from FR (top) and OR (bottom) nominal MIPAS measurements.

Figure 3 .
Figure 3.The upper plot shows an average of 1141 observed (black dots) and simulated (red line) limb radiances in the MWs used for CCl4 10 air in the tropical regions.The CCl 4 distribution shows also a seasonal variability.The intrusion of CCl 4 -poor mesospheric air in the stratosphere during winter, due to the air subsidence induced by the polar vortex, is clearly visible in both polar winters, its effects continuing into early spring and extending into the troposphere.Minimum CCl 4 values are observed in November in the South Pole and in March in the North Pole (November is considered the beginning of spring in the South Pole, whereas Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1163,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2017 c Author(s) 2017.CC-BY 3.0 License.

Figure 4 .
Figure 4. Zonal monthly averages of MIPAS CCl4 profiles.The maps refer to four separate months in different seasons: August 2010 (top left), November 2010 (top right), February 2011 (bottom left) and May 20117 (bottom right).

Figure 5 .
Figure 5.Time evolution of CCl4 at all latitudes, from July 2002 to April 2012.The three maps refer to different pressure levels: 50 hPa (top), 90 hPa (center) and 130 hPa (bottom).

Figure 6
Figure 6 shows the differences between average CCl 4 profiles measured in the Northern Hemisphere (NH) and in the Southern Hemisphere (SH), as a function of latitude and pressure.The time period employed in the calculated averages extends from April 1 st , 2005 to March 31 st , 2012.The estimate of the North-South (N-S) differences is important because for longlived compounds the Inter Hemispheric Gradient (IHG) at the surface is recognized as a qualitative indicator of continuing

5Figure 6 .
Figure 6.Average North-South CCl4 VMR differences versus latitude and pressure.The average period includes MIPAS measurements from April 1 st , 2005 to March 31 st , 2012.

Figures 7 Figure 7 .
Figures7 and 8show the average differences between CCl 4 VMR retrieved from MIPAS/ENVISAT and MIPAS-B both in absolute and relative units.The two figures refer to matching measurements in the FR and the OR phases of the MI-

Figure 8 .
Figure 8. Intercomparison between MIPAS-B and MIPAS/ENVISAT CCl4 VMR.Results for the OR part of the MIPAS mission.
. Phys.Discuss., doi:10.5194/acp-2016-1163,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2017 c Author(s) 2017.CC-BY 3.0 License.algorithm is described in Boone et al. (2005), and the updates for the most recent version of the retrieval, version 3.5, are detailed in Boone et al. (2013).The retrieval algorithm uses a non-linear least-squares global-fitting technique that fits the ACE-FTS observed spectra in given microwindows with forward modelled spectra based on line strengths and line widths from the HITRAN 2004 database (Rothman et al., 2005) (with updates as described by Boone et al. (2013)).Pressure and temperature profiles used in the forward model are the ACE-FTS derived profiles, calculated by fitting CO 2 lines.The spectral window used for CCl 4 retrievals goes from 787.5 to 805.5 cm −1 .

Figure 9 .
Figure 9. Mean CCl4 profile difference between co-located MIPAS and ACE-FTS measurements (red) with standard deviation of the mean (red error bars).The standard deviation of the differences (orange), the estimated total random (green) and total systematic (blue) errors of the difference are also shown The number of co-located pairs for each pressure level is reported on the right side of each graph.Each plot refers to a latitude interval as indicated in the title.

Figure 10 .
Figure 10.CCl4 trend analysis for 20 • S/25 • S at 50 hPa (top left), 55 • S/60 • S at 100 hPa (top right), 25 • N/20 • N at 90 hPa (bottom left) and 50 • N/45 • N at 100 hPa (bottom right).The blue dots are the MIPAS monthly averages and the error bars are the standard deviation of the means.The red curve is the best fitting function VMR(t) and green line is the linear term (trend).The lower part of each plot shows the residuals between the MIPAS monthly averages and the best fitting function VMR(t).The CCl4 trend, its uncertainty and the bias between FR and OR are also indicated in each panel.

Figure 11 .Figure 12 .
Figure 11.CCl4 trends as a function of latitude and pressure.Panel a) absolute trends, b) percentage trends, c) absolute errors, d) percentage errors.Latitudes / pressures with trend error greater than 30% are masked with dashed areas.

Figure 13
Figure 13 compares the time series of ground-based CCl 4 measurements of selected stations (black and orange lines) with MIPAS monthly tropospheric averages (blue dots) in the same latitude bin of the ground station.The two plots refer to ground stations located at tropical (top) and middle (bottom) latitudes.Ground-based measurements do not really show a seasonality, while MIPAS measurements do.The amplitude of the seasonal variations observed by MIPAS increases with latitude.For tropical latitudes MIPAS OR measurements show a positive bias of approximately 15%.Although not focused on tropical 5

Figure 13 .
Figure 13.Comparison between MIPAS (blue dots) and NOAA/AGAGE (black/orange) CCl4 time series.The two plots refer to ground stations located at tropical (top) and middle (bottom) latitudes.The red curve is the fitting model used to derive the trend from MIPAS data, the green line is the linear part of the model itself.The obtained trend values are also shown in the plots.
the authors study the stratospheric lifetime of several species (including CFC-11, CFC-12 and CCl 4 ) using ACE-FTS measurements.Using a CFC-11 lifetime of 45±7 (World Meteorological Organization (WMO), Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1163,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2017 c Author(s) 2017.CC-BY 3CCl4 lifetime calculated using the tracer-tracer linear correlations with CFC-11 and CFC-12 in the lower stratosphere for the entire globe, the Northern and the Southern hemispheres.2011) these authors calculate a CCl 4 global lifetime of 35±11 years.The difference with our results is explained taking into account the different reference CFC-11 lifetimes used: using the same CFC-11 lifetime (World Meteorological Organization (WMO), 2011) we would obtain a CCl 4 lifetime of 40±7 years.As in Brown et al. (2013) we also computed the CCl 4 lifetime separately in the two hemispheres, identifying consistent values: 47(39 -61) years in the NH and 45(38 -59) years in the SH.By contrast, Brown et al. (2013) report very different lifetimes in the two hemispheres (41±9 years in the NH and 21±6

Table 2 .
Overview of MIPAS balloon flights used for intercomparison with MIPAS/ENVISAT multiple spectra recorded at the same elevation angle.MIPAS-B limb scans are typically acquired on a 1.5 km vertical tangent height grid.

Table 3 .
For each ground station the table columns show respectively: site code, site name, site latitude, network name, station-related CCl4 trend, tropospheric MIPAS trend, latitudinal band from which MIPAS data were extracted.