The Umkehr technique, which will be described in Sect. , allows for the low-resolution retrieval of ozone profiles from measurements made by Dobson and Brewer spectrophotometers. TCO and ozone profile measurements with Dobson (and Brewer) spectrophotometers were performed at Arosa (46.82∘ N, 6.95∘ E) from 1926 (and 1988) to 2021 and at Davos from 2012. For a detailed description of the Dobson and Brewer spectrophotometers, we refer to . The progressive relocation of the Dobson and Brewer triads from Arosa to Davos (13 km north of Arosa and 260 m lower in altitude) between 2012 and 2021 is described and analyzed in . Umkehr measurements have been performed under clear-sky and low cloud cover conditions twice a day since 1956 by Dobson spectrophotometers (Dobson D015 since 1956 and then Dobson D051 since 1988) and four to six times per month by Dobson D101 since 1988 and by Dobson D062 since 1998. Dobson D051 performs fully automated Umkehr measurements since 1988. The Dobson Umkehr measurements have been complemented by Brewer Umkehr measurements since 1988 with Brewer B040 and since 2005 with Brewers B072 and B156. See Table  for a summary of the time ranges and time resolutions of the six Arosa/Davos spectrophotometers. Ozone profile Umkehr measurements were initiated in 1956 at Arosa and were continued from 2021 at Davos, Switzerland. They compose the longest continuous Umkehr measurement time series worldwide .

At Arosa, the Dobson D051 sat on a turntable in a conditioned hut maintained at 25–28 ∘C. An aperture in the roof, which opened and closed according to solar zenith angle (SZA) and weather conditions, allowed for zenith measurements. The continuous and automated measurements (2 min cycle) are interpolated to 12 nominal SZAs, and profiles are retrieved from the ground to 50 km using the optimal estimation method (OEM) implemented in . Manual Umkehr measurement started in 1968 with the Dobson D101 and in 1992 with the Dobson D062 as redundant measurements to check the stability of Dobson D051. These have been made two to three times each month since 1988 and 1998 in favorable weather conditions. The Dobson D062 and the Dobson D101 were automated in 2012 and in 2013, respectively . They have been located since 2021 in Davos in a common air-conditioned container side by side with the Dobson D051 and measure Umkehr curves through a quartz dome. While the Dobson D051 was dedicated exclusively to Umkehr measurement until February 2013, the present setup allows for both direct sun and zenith Umkehr measurements with the three Dobsons. The Arosa/Davos Dobson instruments are regularly calibrated against the two European regional secondary reference Dobson instruments D064 from the Hohenpeissenberg Observatory (MOHp, Germany) and D074 from the Solar and Ozone Observatory in Hradec Králové (SOO-HK, Czech Republic) Fig. 3.

The Brewer triad consists of two Brewer Mark II single-monochromator instruments, the Brewer B040 and the Brewer B072, and one Brewer Mark III double-monochromator instrument, the Brewer B156. The three instruments measure daily in Umkehr mode when the sun is at the 12 nominal SZAs. Since the operation of the first Brewer at Arosa in 1988, biennial calibrations have been carried out Fig. 1 towards the traveling reference instrument Brewer B017 and, since 2008, towards the traveling reference instrument Brewer B185. The instruments of the Brewer triad underwent very few technical interventions and are in good agreement with the traveling references (TCO deviations ≤ 1 %; ). In particular, no technical issues are reported around 2011–2013 and 2018, which are data record periods considered in the frame of the Dobson D051 homogenization. However, sporadic instabilities in the Brewer B072 data record have been observed, while no particular technical issues have been detected by the intercomparison procedures. The Dobson D051, the Brewer B072 and the Brewer B156 were simultaneously relocalized from Arosa to Davos in September 2018 but with an effect on the TCO level within the instrumental noise .

The Microwave Limb Sounder (MLS) is a microwave limb-sounding radiometer on board the Aura Earth-observing satellite, launched in July 2004. Ozone profiles are retrieved from Aura MLS radiance measurements at 240 GHz. Details about the instrument can be found in . Ozone profiles from the version 4.2 dataset are given on 55 pressure levels from 1000 to 1 × 10-5 hPa . However, the useful vertical range for Aura MLS ozone leads us to only consider Aura MLS data from 10 to 75 km (in this range, the Aura MLS vertical resolution is about 2.5–4 km) for Aura MLS overpasses above Arosa (±3∘ in latitude and ±5∘ in longitude). These ozone profiles are interpolated on the Umkehr pressure levels pi and converted to DU following : 1XDU=C⋅X‾⋅(pi-pi-1), with C = 0.00079 DUhPa-1ppbv-1 and X‾ the ozone mean volume mixing ratio (VMR) in parts per billion per volume (ppbv). Approximative heights are given as in .

The Arosa/Davos Umkehr time series is composed of Dobson D015 measurements from 1956 to 1988 and Dobson D051 since then. The quality of the homogenization of the Dobson D015 to Dobson D051 transition has been ensured by 1 year of parallel measurements (1988), allowing for an adaptation of the D015 N values to the D051 N values. For each SZA, the 1988 mean difference between the D051 and the D015 N values has been added to the D015 values. The 1956–1987 ozone profiles have then been retrieved from the Dobson D015 corrected N values. No statistical correction has been performed on the D015 ozone dataset.

We report here about the complete homogenization of the 1988–2020 Umkehr Dobson D051 time series by comparison to the datasets of the five collocated instruments (two Dobson and three Brewer spectrophotometers) on the N value level. The purpose is to detect common anomalies in the difference between Dobson D051 and each of the redundant measurements and to correct the Dobson D051 time series accordingly. However, a correction is only applied if it correlates with a technical issue reported in the metadata. If we cannot see any indication in the metadata for an instrumental drift, no correction is applied.

Figure  shows the time series of monthly mean ozone profile differences between Dobson D051 and the five collocated spectrophotometers. Only simultaneous measurements, not flagged for bad weather conditions, volcanic eruptions, and number of iterations, are considered. The relative differences of the anomalies lie within ±15 %. The comparisons with the Brewer instruments show a seasonal cycle with differences slightly bigger in summer than in winter (not shown; DL6: -2 % in winter and +2 % in summer). A similar behavior has been found by when comparing TCO from Dobsons to Brewers. Note that the annual cycle is not visible on the representation of deseasonalized anomalies as in Fig.  and that we consider changes when they are larger than the standard deviation of the Brewer Dobson differences.

If we focus on the post-2000 period, where several collocated and redundant measurements are available, systematic anomalies of the Dobson D051 are noticed (periods in black frames in Fig. ):

Before 2003 for the altitude range below 30 km, the Dobson D051 ozone values are higher than the values measured by the collocated instruments below 20 km and lower between 20 and 30 km.

The comparison of Dobson D051 with the collocated Dobsons around 2014 and after 2018 is to be taken with caution due to the very limited number of measurements of Dobson D051 in 2014 and of Dobson D062 and Dobson D101 during these periods. Around 2014 (technical and staff transition period), many data are missing or have to be flagged because of roof opening issues. Since 2018, the Umkehr measurements by Dobson D062 and Dobson D101 have been drastically reduced as priority has been given to total ozone measurements.

Table  summarizes the Dobson D051 problematic periods, the technical issue reported at these periods, and the time ranges and redundant datasets used for the offsets determination.

When systematic for each pair of instruments and if related to an instrumental issue, the detected Dobson D051 problematic periods are shifted according to the mean difference with the three Brewer or the two Dobson datasets before and after the problematic periods (periods of 2 years are considered). The homogenization is performed on the raw data level (N values), and the ozone profiles are then retrieved from the corrected N values. The Dobson D051 and the Brewers stay independent from each other as one is not corrected to fit the ozone values of the others. Only the mean variation of the Brewers datasets during 2 years before and after an anomaly (Brewers data records do not suffer from anomalies during these periods) is replicated on the same 4 years of Dobson D051, allowing the long-term ozone variations to stay independent.

For each period that requires a correction (see Table ), we apply a SZA-dependent offset to the N values which is constant over the period to be corrected. The offset is calculated such that the difference averaged over the period and over the reference instruments (two Dobsons in 2003 or three Brewers after 2011) matches the difference averaged over 2 years before and 2 years after the period and over all reference instruments (see Fig. ): 2ΔSZA=mean(Δ1SZA,Δ3SZA)-Δ2SZA3N2SZAcorr=N2SZA-ΔSZA.

ΔSZA is the offset between the three Brewer mean N values and the Dobson D051 N values for each SZA, and Δ1SZA and Δ3SZA are the difference between the three Brewers mean N values and the Dobson D051 N values before (period P1) and after (period P3) the Dobson D051 problematic period (period P2). All values are averaged over 2-year periods. N2SZAcorr is the corrected N value in period P2.

In case of a step in the time series (e.g., in July 2003 and in May 2018), the period P2 does not exist and should not be considered in Fig. . The corrected N value N2SZAcorr of period P1 is then obtained following Eqs. () and (). 4ΔSZA=Δ1SZA-Δ3SZA5N1SZAcorr=N1SZA-ΔSZA

In parallel but in a separate work, a homogenization and a correction for the stray light effect of the same Dobson dataset have been performed by NOAA . They use the comparison of the Dobson D051 dataset with the M2GMI model on the N value level when the MCH homogenization uses the comparison with N values of the collocated instruments. A summary of the homogenization method is presented here; for details on the method and for the description of the stray light correction, we refer to .

The NASA Global Modeling Initiative Chemistry Transport Model (GMI CTM) is a full general circulation model that is driven by MERRA2 meteorological reanalysis through the replay method . The simulation of the meteorological fields in the M2GMI model is continuously referenced against the MERRA-2 winds, temperature and surface pressure fields . For the NOAA homogenization process, the M2GMI ozone and temperature profiles are selected for the Arosa station location. The simulated temperature profile is used for accounting for the temperature dependence of the ozone cross section and allows for the model to better fit to the day-to-day variability of the N values. The Umkehr retrieval forward model uses the M2GMI profiles to simulate Umkehr N values for an idealized Dobson instrument that does not have a stray light interferences. For each SZA, differences between simulated (idealized) and measured (instrument specific) Umkehr N values are averaged over the time between two consecutive calibrations (performed at each Dobson intercomparison campaign) of the Dobson D051 to create an empirical correction that accounts for the stray light of the D051 instrument. An iterative modification of the N value correction is further performed for optimization of the stray light correction, adding a constant offset correction to the Umkehr dataset. This results in a reduced bias to other ozone records in the upper stratosphere but, as a constant offset, does not have any impact on the trends. While the first iteration of the homogenization removes artificial steps in the Umkehr ozone profile records, the iterative part reduces the bias relative to other ozone observing systems.

The NOAA homogenized Dobson D051 dataset has been compared to satellite data records including Aura MLS in . The agreement is within ±-5 % in the upper and middle stratosphere, and larger biases (up to 10 %) are found in the lower stratosphere.

The NOAA homogenization has been developed to remove artificial steps in the Umkehr ozone profile records and to reduce the bias relative to other ozone observing systems. The MCH homogenization approach is different in that the homogenization process aims to remove artificial steps in the Dobson D051 Umkehr profiles record while maintaining the constant offset between the datasets, thus ensuring the independence of the Dobson D051 ozone values towards the collocated instruments datasets.

Both homogenization processes provide correction offsets on the N value level and ozone profiles retrieved from the corrected N curves. We compare first the time series of the N value correction offsets. Then the homogenized ozone profile time series are considered by comparing the time series of their difference to Aura MLS.

Figure  shows the time series of the N value correction as a function of SZA as determined by the NOAA homogenization (Fig. a) and by the MCH homogenization (Fig. b, this study). For comparison purposes, the NOAA correction values were offset with their mean difference after 2018.

The main differences between the two homogenization results are the variability of the corrections values and the correction of the volcanic eruptions periods. The seasonal variability of the NOAA N values comes from the correction of observed N values for the stray light effect. Indeed, the stray light contribution varies with SZA and is proportional to the total column ozone value . For the same SZA, the amount of correction is different for each monthly mean value of the time series in proportion to the seasonal changes in total column ozone (Fig. a). This is not corrected for in the MCH N value homogenization. The years around 1982 and 1992 are periods of volcanic eruptions (El Chichón and Pinatubo) which are corrected by the NOAA homogenization but not considered in the MCH homogenization as the Umkehr retrieval does not account for the change in atmospheric scattering due to aerosol injection . For the 1988 to 2003 period, both homogenization results differ for the 77–83∘ SZAs. Otherwise, the correction amplitudes are similar, and their occurrences coincide within a few months in January 1988, in July 2003 and in April 2011 to 2013; this is remarkable given the differences in the detection method. Note that the 2010 6-month step has been chosen to be left uncorrected in the MCH homogenization due to the absence of confirmed technical issue at that time. In 2017/2018, the start date considered for the NOAA homogenization is January 2017, while the start date considered by the MCH homogenization is May 2018 with the probable effects of the wedge steel band replacement on the measurements.

However, while both corrections of the N values look similar, small differences in the N curve shapes can lead to larger differences in the ozone profiles due to the nonlinear relationship between the N values and the ozone values (see the two N curves and ozone profiles in Fig.  for an example).

In order to evaluate the effects of both homogenization on the Dobson D051 time series, monthly mean relative difference to Aura MLS data record are plotted in Fig.  for two altitude levels, i.e., DL5 (25 km) in the middle stratosphere and DL8 (40 km) in the upper stratosphere. The relative difference of the Brewer B040 time series is also shown for the same layers.

The Brewer B040 relative difference shows a constant offset to Aura MLS but clear anomalies in 2012 and 2013 in DL5 (Fig. a). The Dobson D051 homogenized by NOAA shows a very good accordance with Aura MLS both in DL5 and DL8. The small mean bias is a result of the NOAA optimization of the stray light correction. Therefore, it is not the magnitude of the bias between the homogenized dataset and Aura MLS but its variation (the bias should be constant) which should be considered here. No clear offset in the difference to Aura MLS between the NOAA and the MCH homogenized record is reported in DL5. The variability of the differences to Aura MLS of each dataset looks higher after 2010, while the mean values are constant. However, the slight underestimation of the MCH homogenization since 2017 seems to match the Brewer B040 difference to Aura MLS in DL5 (Fig. a). After 2017, the relative difference to Aura MLS of D051 homogenized by MCH and of the collocated B040 is within -5 % to -10 % while the D051 homogenized by NOAA lies within -2 % of Aura MLS.

A clear correction of the 2011–2013 period is visible in DL8 (Fig. b). Except for the respective MCH and NOAA homogenized dataset mean offsets to Aura MLS, a slight overestimation of the NOAA homogenization is visible in 2012 and 2013. However, the Brewer B040 relative difference to Aura MLS is also slightly smaller during this time range, when the Brewer instrument had not undergone any technical interventions. This is particularly visible on the anomalies time series of B040 in Fig. c. As the MCH homogenization relies on the Brewer collocated datasets, it allows the local variability of the ozone DL8 content to be taken into account, which the M2GMI model, basis for the NOAA homogenization, probably does not consider. As the atmospheric processes are more homogenized in the stratosphere than in the troposphere, the M2GMI ozone profiles should be representative of stratospheric ozone variability. Nevertheless, it is possible that other atmospheric interferences (i.e., aerosols) can impact the Dobson readings of zenith sky radiance which would also impact Brewer observations but might not be fully included in the M2GMI simulations.

Due to the occurrence of an anomaly in 2018, which is particularly visible in DL8 for all datasets (Fig. c), the last correction applied to the dataset by the NOAA and the MCH homogenization differs.

As the MCH homogenization considers a step correction in May 2018, the ozone increase during the 2018 anomaly is accounted for in the mean difference of the D051 dataset to the Brewers datasets of the pre- and post-step periods. As a result, the calculated offset is small. The NOAA homogenization method detects a change in the Umkehr ozone with respect to the M2GMI record that starts a year earlier, in 2017. The ozone increase during the 2018 anomaly is only accounted for in the mean difference to M2GMI of the post-step period of the D051 dataset. Moreover, this post-step difference is overestimated as M2GMI does not seem to simulate any significant anomaly at that period. As a result, the calculated offset, applied in 2017, is probably overestimated.

Now that the Dobson D051 is fully homogenized, vertically resolved long-term trends can be estimated with limited influence of instrumental artifacts.

Trends are estimated by fitting a multilinear regression function to the monthly mean ozone time series considering two piecewise linear trends (PWLT) starting in 1970 and in 1998. Trend profiles are obtained by considering one independent monthly mean time series for each pressure level. The results are given as a difference in DU to the 1970–1980 and of the 2000–2010 means. The explanatory variables represent sources of geophysical variability with known influence on stratospheric ozone, including the quasi-biennial oscillation (from https://www.geo.fu-berlin.de/met/ag/strat/produkte/qbo/index.html, last access: 26 July 2022) (QBO) at 30 and 10 hPa, the 10.7 cm solar radio flux describing the 11-year solar cycle (from https://www.iup.uni-bremen.de/UVSAT/Datasets/mgii, last access: 26 July 2022) (SOL), the El Niño–Southern Oscillation (from http://www.esrl.noaa.gov/psd/enso/mei/, last access: 26 July 2022) (ENSO), the North Atlantic Oscillation (from https://climatedataguide.ucar.edu/climate-data/hurrell-north-atlantic-oscillation-nao-index-station-based, last access: 28 November 2021) (NAO), the stratospheric aerosol optical depth (from https://asdc.larc.nasa.gov/project/GloSSAC/GloSSAC_1.0, last access: 28 Novemeber 2021) (SAOD) and Fourier components representing the seasonal cycle (annual and semi-annual variations). All data points are considered with equal weights, and the uncertainty of the fit parameters is estimated from the regression residuals. Residual autocorrelations are accounted for by applying a Cochrane–Orcutt transformation to the model .

Dynamical linear modeling allows for the determination of a nonlinear time-varying trend from a monthly mean time series. This is a Bayesian approach regression which fits the data time series for a nonlinear time-varying trend, regression coefficients from explanatory variables, and seasonal and annual modes, considering their uncertainties and an autoregressive component. The trend is allowed to smoothly vary in time, and its degree of nonlinearity is inferred from the data, as well as the turnaround period. We use the code by , which is a Python implementation of the formalism introduced by , and we refer to these publications for a detailed description of the DLM principles. The model used considers standard regression components, allows for a variability of the sinusoidal seasonal modes and includes the autoregressive (AR1) correlation process with variance and correlation coefficient as free parameters in the regression. The same five explanatory variables as in the MLR are used in the trend estimate: QBO at 30 and 10 hPa, SOL, ENSO and SAOD NH values. The estimation of the posterior uncertainty distribution is performed with the Markov chain Monte Carlo (MCMC) method and considers the uncertainties on the regression components, on the seasonal cycle, on the autoregressive correlation and on the nonlinearity of the trend. Note that only statistical uncertainties are given in the paper, which allows the significance of the trends to be determined. In order to check the agreement of trends derived form different datasets, uncertainties including a term accounting for remaining steps and for inhomogeneities in the dataset should be considered.

Figure  shows the long-term trend estimates from the MCH homogenized Dobson D051 dataset by DLM (in blue with ± 2σ uncertainty shaded area) and by MLR (PWLT, in black with ± 2σ uncertainty shaded area) for the same explanatory variables at three altitude levels. The blue shaded areas show the non-constant 2σ uncertainties in Dobson units per year estimated by the DLM. By analogy, for the MLR, the grey shaded areas report the uncertainty in Dobson units per year calculated from the constant 2σ offset trend uncertainty in Dobson units per decade.

Overall trends are similar but differ over short timescales because of their representation of the nonlinearity of the changes in the data record. The advantage of DLM lies in the estimation of a smoothly varying trend without assuming any shape. The inflection year depends on the method: while the inflection point is fixed by the MLR PWLT (1998 in this case; see ), the inflection year is retrieved by the DLM and results in year 2002 for the Dobson D051 dataset above 28 km. The maximum of the ΔO3 (ozone difference) between 1998–2020 KDE (kernel density estimation) should be compared to the linear trend value over the same time period (22 years), while the 95 % level of significance, represented by the fraction of the KDE above/below zero, slightly differs from the MLR uncertainty estimates. In the lower stratosphere, for DL4 (Fig. a and d), the post-1998 MLR trend values are -0.33 ± 1.35 %perdecade. The DLM KDE shows a maximum at -1.64 DU and a full width at half maximum (FWHM) (i.e., 2.4σ for normal distribution) of 1.81 DU, which means a mean trend of -1.09 ± 1.10 %perdecade. The MLR estimate is nonsignificantly different from zero at the 95 % confidence level, while the DLM estimate is negative and barely significant at the 95 % level. In the middle stratosphere, for DL5 (Fig. b and e), the post-1998 MLR trend values are -0.52 ± 0.86 %perdecade. The DLM KDE shows a maximum at -1.49 DU and a FWHM of 1.18 DU, which means a mean trend of -1.09 ± 0.78 %perdecade. The MLR estimate is nonsignificantly different from zero at the 95 % confidence level, while the DLM estimate is significantly negative at the 95 % level. In the upper stratosphere, for DL8 (Fig. c and f), the post-1998 MLR trend values are +4.84 ± 1.40 %perdecade, and the DLM KDE shows a maximum at +0.80 DU and a FWHM of 0.41 DU, which means a mean trend of +3.59 ± 1.69 %perdecade. Both are significantly positive at the 95 % confidence level. The estimated post-1998 MLR trends are in agreement with the vertically resolved trends reported in the literature , and the post-1998 MLR and DLM estimated trends are in agreement within their uncertainties. The lower and middle stratospheric trends differ in their significance though. In case of high annual variability, a DLM trend estimate in percent per decade may be significant, while a MLR trend estimate may be nonsignificant for the same considered period. Note that the given DLM trend value in percent per decade is an average of the percentage change per year. The regressions (resulting trends and their uncertainties) are influenced by outliers , but trends estimated by DLM regression change each year. Hence, outliers only influence a limited portion of the DLM trend time series, influencing only the associated trend uncertainties.

Figure a and b show the DLM trend estimates derived from the Dobson D051 record as homogenized by MCH (Fig. a) and by NOAA (Fig. b). The trend values are given in percentchangeperyear for each altitude level between 10 and 50 km. Positive and negative trends are shown with varying intensities of red and blue, respectively. The grey lines indicate trend estimates nonsignificantly different from zero at the 95 % confidence level.

The upper stratospheric (DL7–10, 10–1 hPa, 35–50 km) trend estimates are significantly negative between 1965 and 1997 in Fig. a and before 1997 in Fig. b. The mean negative trend estimates are -5 %perdecade (mean value of the 1965–1997 upper stratospheric trends). Both records then show a transition period until 2003, with nonsignificant upper stratospheric trend estimates. The post-2003 upper stratospheric trends are significant and positive, up to 2020 for the MCH homogenized Dobson D051 record and until 2013 for the NOAA homogenized Dobson D051 record. The mean positive upper stratospheric trends are 3.6 %perdecade in Fig. a (mean value of the 2003–2013 upper stratospheric trends) and 2.1 %perdecade in Fig. b (mean value of the 2003–2013 upper stratospheric trends). Note that due to the large AKs of the Umkehr measurement, the ozone and trend information in DL8 and DL9 is not independent of each other. In the middle stratosphere (DL5 and 6, 24–32 km), both homogenized records show a negative trend in DL5, persistent and significantly different from zero at the 95 % confidence level since 2012 for the MCH homogenized Dobson D051 data record but slightly positive between 2002 and 2010 and nonsignificantly different from zero at the 95 % confidence level for the NOAA homogenized data record. In the lower stratosphere (LS; DL3 and 4, 14–24 km), the DL3 and DL4 trend estimates are nonsignificantly negative before 1996 but significantly negative between 2008 and 2018 in DL4 for the MCH homogenized data record and nonsignificantly negative for the NOAA homogenized Dobson D051 record.

Again due to the AK width of the Umkehr profiles, the ozone content information of DL2 partly overpasses the lower stratosphere as usually defined (see representation of shaded areas in Fig. b). The same consideration is true for the DL6 in the middle stratosphere. The lower part of the lower stratosphere and the upper part of the middle stratosphere trends may be aliased by upper tropospheric and upper stratospheric information, respectively.

Post-2000 trends have been estimated on the three Dobson and the three Brewer MCH Umkehr data records. The trend estimates of one of the Dobsons (D051), one of the Brewers (B040) and Aura MLS are represented in Fig. a–c in percent change per year for each altitude level between 10 and 50 km.

The post-2000 trends show similar features for the Dobson and Brewer spectrophotometers:

There is a positive trend of 0.2 to 0.5 %yr-1 above 35 km, significant for Dobson D051 (and for Dobson D062 and Brewer B156 not shown) but lower and therefore nonsignificantly different from zero at the 95 % level of confidence for Brewer B040 and Dobson D101. Despite differences in the trend estimate intensities, an overall picture of a upper stratospheric positive trend after 2002 is shown.

Significant upper stratospheric positive trends are estimated on the Aura MLS satellite data record (Fig. c) but have been nonsignificant since 2013. Signs of negative trends in the lower altitudes are also observed although not significant: DLM trend estimates have been persistently negative in the middle stratosphere and negative in the lower stratosphere since 2012.