In this work, we use data from four missions conducted with the High Altitude and Long Range Research Aircraft (HALO). HALO is a Gulfstream V (GV) aircraft and can reach altitudes of up to 15 km. The first mission is the TACTS (Transport and Composition in the Upper Troposphere/Lowermost Stratosphere) mission, conducted in August and September 2012 with flights covering an area from the Cabo Verde islands to the Norwegian archipelago of Spitsbergen and over Europe and the Atlantic Ocean. The base of all flights was Oberpfaffenhofen (Germany) (Fig. , blue tracks). The second mission was PGS, consisting of three sub-missions: POLSTRACC (Polar Stratosphere in a Changing Climate), GW-LCYCLE (Investigation of the Life cycle of gravity waves), and SALSA (Seasonality of Air mass transport and origin in the Lowermost Stratosphere). The mission took place from December 2015 to March 2016 with flights mainly in the Arctic and covering Greenland, the North Atlantic, and Europe. Flights were conducted from Oberpfaffenhofen (Germany) and from Kiruna (Sweden) (Fig. , purple tracks) . The third mission was the WISE (Wave-driven ISentropic Exchange) mission between September and October 2017. Flights were conducted mainly from Shannon (Ireland), covering an area above the Atlantic Ocean and western Europe (Fig. , green tracks). Finally, the SouthTRAC (Southern Hemisphere Transport, Dynamics, and Chemistry) mission took place from September to November 2019. It is the only one of the four HALO missions that covers the SH. In addition to the scientific transfer flights, which departed from Oberpfaffenhofen (Germany) via the Cabo Verde islands to South America, all other flights took place from Rio Grande (Argentina). Flights of the SouthTRAC mission cover the southern Pacific and southern Atlantic oceans near South America and Antarctica (Fig. , orange tracks).

Data from the Gas chromatograph for Observational Studies using Tracers (GhOST) in situ instrument were used in this analysis. The instrument has two channels. The first channel couples an isothermally operated gas chromatograph (GC) with an electron capture detection (ECD) (GhOST-ECD), and the second channel couples a temperature-programmed GC with a quadrupole mass spectrometer (MS) (GhOST-MS) (see and references therein). For the SouthTRAC campaign, the GhOST-MS ionization mode was changed from negative chemical ionization (NCI) to electron impact ionization (EI) to record broader mass spectra, leading to different detection limits for the Br-VSLSs compared to previous campaigns. In this work, only measurements of the MS channels are used. The measurements of CH2Br2 and CHBr3 are on the NOAA-2003 scale, thus consistent with NOAA/ESRL observations.

The HIPPO mission measured cross sections of trace gases over the Pacific Basin and North American mainland (170–80∘ W) (Fig.  black tracks), covering a latitudinal range from the North Pole (85∘ N) to the coastal region of Antarctica (65∘ S) . The mission was split into seasonal segmented deployments, which took place in January 2009 (HIPPO-1), October to November 2009 (HIPPO-2), March to April 2010 (HIPPO-3), June to July 2011 (HIPPO-4), and August to September 2011 (HIPPO-5). The platform used for the observations was the NSF/NCAR High-performance Instrumented Airborne Platform for Environmental Research (HIAPER) Gulfstream V (GV) aircraft.

Data from two whole air samplers (WASs) were combined in this analysis. The University of Miami operated the Advanced Whole Air Sampler (AWAS), storing air samples in pressurized stainless steel canisters . The second sampler operated during HIPPO was the NOAA Whole Air Sampler (NWAS), which stores samples in pressurized glass flasks. Subsequently, the samples were analysed using ground-based laboratory GC–MS (gas chromatograph–mass spectrometer) systems. Results from both laboratories were provided on a scale consistent with NOAA/ESRL ground-based station results see.

The ATom mission was split into four parts, which took place in July to August 2016 (ATom-1), January to February 2017 (ATom-2), September to October 2017 (ATom-3), and April to May 2018 (ATom-4). Thus, all seasons were covered within a 2-year time period. In each season, flights started and ended in Palmdale (California, USA) with a route to the western Arctic, south to the South Pacific, east to the Atlantic, and north to Greenland, returning across central North America (Fig. , red tracks). The platform for the ATom mission was the NASA DC-8 aircraft, which is capable of reaching an altitude of around 12 km .

As with the HIPPO mission, data from two WASs were used for this analysis. The University of Carlifornia – Irvine (UC-Irvine) research group operated one whole air sampler (hereafter referred to as WAS), storing air in stainless steel canisters. Samples were analysed in the laboratory using GC with flame ionization detection (FID), ECD, and MS . In addition, the NOAA/GML's Programmable Flask Package Whole Air Sampler (PFP) was operated using glass flasks. The air samples were analysed at the NOAA's Global Monitoring Division laboratory for trace gases by GC-MS and at the Institute of Arctic and Alpine Research (INSTAAR) Stable Isotope Lab for isotopes of methane. Beside the two WASs, the NCAR Trace Organic Gas Analyzer (TOGA) was operated during ATom. TOGA is an in situ instrument, combining a GC with an MS , similar to the GhOST-MS. used ATom-2 observations among others to improve estimates of short-lived halocarbon emissions during summer from the Southern Ocean using airborne observations. The two whole air samplers and the TOGA instrument shared approximately half of the sampling period, generally presenting a good correlation and consistency in mole fractions for CH2Br2 and CHBr3 .

The TOMCAT (Toulouse Off-line Model of Chemistry And Transport) is an Eulerian offline three-dimensional chemistry transport model (CTM) . The model uses a hybrid vertical sigma–pressure coordinate (σ–p) with 60 vertical levels from the ground up to around 60 km. The horizontal resolution was set to 2.8∘×2.8∘ (latitude × longitude). The CTM is forced by meteorological fields (winds, temperature, and humidity) taken from the European Centre for Medium-Range Weather Forecasts ERA5 reanalysis . The internal model step was 30 min and monthly means of the tracers are generated for this study. A similar setup was previously used to study NH Br-VSLSs in , beside using ERA-Interim data instead of ERA5. In addition, the configuration used here reads an offline monthly varying climatological OH concentration field, developed for the TransCom-CH4 project . In this study, the VSLS emission scenario of , which includes monthly variability in emissions, was used with TOMCAT. Model output is available for the period from 2009 to 2019.

CAM-Chem (Community Atmosphere Model with Chemistry, version 4) is a three-dimensional chemistry climate model (CCM) and a component of the NCAR Community Earth System Model (CESM) . The WACCM (Whole Atmosphere Community Climate Model) physics module for the stratosphere is included, and it uses the chemical mechanism of MOZART (Model for OZone and Related chemical Tracers) with different possibilities of complexity for tropospheric and stratospheric chemistry. The model includes a detailed treatment of tropospheric Br-VSLS sources and chemistry described in and . The horizontal resolution was set to 0.96∘×1.25∘ (latitude × longitude) and 56 hybrid vertical levels from the surface to around 40 km. The model setup is similar to the one used in , using the NASA Goddard Global Modeling and Assimilation Office (GMAO) GEOS5-generated meteorology. The model step was 5 min but monthly means of the tracers are used for this study. As with the TOMCAT model, the monthly varying emission scenario of was used, with fixed emissions of the VSLSs during the whole modelling period (available from 2009–2019).

Aeroplanes modified for scientific observations are equipped with on-board instruments to gather meteorological and aircraft parameters along the flight tracks. In addition, local tropopause information along the flight tracks as well as equivalent latitude were derived using the Chemical Lagrangian Model of the Stratosphere (CLaMS) e.g. with underlying ECMWF reanalysis. For the SouthTRAC, PGS, HIPPO, and ATom mission, the underlying meteorological field are taken from the ERA5 reanalysis, whereas for TACTS and WISE, the underlying fields are from the ERA-Interim reanalysis .

For this work, the potential vorticity (PV)-based dynamical tropopause is used e.g.. The commonly used value of 2 PVU (potential vorticity unit) was used for the dynamical tropopause; this condition was replaced by the potential temperature level of 380 K in the tropics when the 2 PVU level is above e.g.. We additionally used the same PV-based climatological tropopause information as in , which is based on the ERA-Interim reanalysis.

In the course of this work, we progressively move from a more global view of the distribution of the tracers to a more detailed view of the UTLS in the midlatitudes of both hemispheres. For the global view (Sect. 4.2), we use latitude as the horizontal coordinate and pressure as the vertical coordinate and thus get a detailed perspective of the troposphere from the ground up to the tropopause and the LMS. As we then look more closely at the area around the tropopause and into the stratosphere, we change the vertical coordinate using potential temperature (Θ) and potential temperature difference to the local tropopause (ΔΘ) instead of pressure (starting from Sect. 4.3). Transport in the free atmosphere is predominantly isentropic, making Θ a very useful coordinate. Furthermore, Θ allows for better vertical resolution as it increases more rapidly with height in stable layers. Finally, the focus moves towards the UTLS of the midlatitudes in the course of the analysis (Sect. 4.4). We switch from the examination in latitude only to a combined coordinate already used in and . Latitude is used for tropospheric observations, whereas equivalent latitude is used for stratospheric ones. The equivalent latitude is a commonly used horizontal coordinate for studying tracers in the stratosphere and assigns PV to latitude based on the area (of the polar cap) enclosed by the specific isopleth of PV on a given potential temperature contour . The combined coordinate is referred to as equivalent latitude*. In all sub-analyses (Sects. 4.2–4.4), the observations are compared with the model data. The model results are only used for the years and months for which observations are available and have been zonally averaged consistent with.

We combined the measurements from the different missions, leaving aside that the Br-VSLSs may have shown a weak positive trend e.g. tropical mean 0.017 ± 0.012 ppt Br per decade for 1979–2013 from. Observations and model results were split by season (DJF: December, January, February; MAM: March, April, May; JJA: June, July, August; SON: September, October, November). The data are binned in 10∘ latitude intervals from 90∘ S to 90∘ N. In the vertical we have binned the data between 1000 and 50 hPa into 20 bins. The bin size decreases logarithmically with increasing altitude and thus lower pressure. Thus, the size of the bins varies from about 180 hPa near the ground to 8.5 hPa in the lower stratosphere, which corresponds to an altitude resolution of about 1.3 km.

Figures  and  show the distributions of CH2Br2 and CHBr3 for the observations (a–d), the TOMCAT model (e–h), the CAM-Chem model (i–l), and the differences between the respective model and the observations (m–t). The merged observational data set allows for a comprehensive representation of the tropospheric distribution except for the southern high latitudes (greater than 70∘ S) in summer and winter. Furthermore, the LMS of both hemispheres is much better covered by observations in spring and autumn.

The distributions of CH2Br2 in the troposphere (Fig. a–d) show a general increase in mixing ratios with increasing latitude, which is most pronounced in hemispheric winter. Increased mixing ratios almost reach the tropopause for this season. The tropospheric distributions show a clear seasonality in both hemispheres with the largest values observed in hemispheric winter and the smallest values in hemispheric summer in the lower troposphere. There is a slight asymmetry towards generally higher mixing ratios in the NH, particularly for the 70∘ N bin. The NH has more coastal regions, which are assumed to be one of the main sources of Br-VSLSs, which could explain the asymmetry of tropospheric mixing ratios. A rather striking difference between the NH and SH is observed in the LMS in autumn and spring. While the distribution of CH2Br2 in hemispheric spring is quite similar in both hemispheres, the distribution in hemispheric autumn differs with smaller values in the SH compared to the NH. Mixing ratios above the subtropical and extratropical tropopause are up to 0.3 ppt smaller in the SH. High mixing ratios of tropospheric tracers in the LMS observations during NH autumn have been explained by a strong influx of tropospheric air during NH summer and autumn (flushing of the LMS) e.g.. It could be argued that this is an indication of the different strength of tropospheric air mixing into the LMS of the two hemispheres. The subtropical jet acts as a transport barrier and diagnosed a hemispheric asymmetry of the subtropical jet with a most pronounced weakening in the NH summer (see Fig. 1 and references in ).

Even though SH observations are available for all seasons from the different missions, the SH database remains much smaller than the NH database. Unfortunately, MAM measurements during ATom-4 show quite large differences between the results from TOGA and both whole air samplers (WAS and PFP) but only for observations of the SH LMS. Figure  displays the altitude–latitude cross section of the observations for MAM, taking all observations (a) and thus being similar to Fig. b as well as using all observations but only TOGA measurements from ATom (b) and using all observations but only the whole air samplers for ATom (c). Although there is little difference in the rest of the atmosphere due to the use of TOGA or WAS/PFP, the difference in the SH LMS during MAM is clear (see Fig.  observations inside red rectangles). Using the TOGA data (Fig. b), values are larger in the range of around 0.4–0.8 ppt, which would suggest a flushing of the LMS similar to that in northern hemispheric autumn e.g.. Values are much smaller when using WAS and PFP data (Fig. c) below around 0.4 ppt. This would indicate a strong isolation of the SH lower stratosphere. Indeed, investigated a strong seasonal cycle in the strength of the barriers at the subtropical jet, where in MAM in the SH, mixing follows a mostly zonal pattern and the subtropical jet acts as a barrier. They further stated that observed mixing is of the greatest magnitude in the NH in any season. The representation of the southern hemispheric UTLS in MAM is based on fewer observations than, e.g., southern hemispheric UTLS in spring where the SouthTRAC campaign took place. The SouthTRAC campaign contributes to a substantial portion of the spring UTLS observations in the SH. For a more meaningful result especially in SH autumn, but also in winter and summer, further measurements are necessary and should be a focus of future campaigns.

The model results show a general good agreement to the observations in the annual mean. Positive or negative bias to the observations are not very pronounced or consistent. Instead, negative or positive bias is dependent on season and latitude. For the case of CH2Br2, the largest lower-tropospheric values in the NH observations are in winter, whereas models show the largest values in spring. This may arise from a possible incorrect seasonal representation in the emission scenario. Furthermore, TOMCAT values in the free troposphere are larger than in CAM-Chem, despite using the same emission scenario. Thus, overestimation in MAM in the NH is larger within TOMCAT with differences to observations of up to about 0.4 ppt. Both models underestimate NH high-latitude values of CH2Br2 in SON by up to 0.3 ppt. showed a comparison of different models and ground-based stations in which the models do not reproduce the seasonality at coastal stations such as Mace Head (Ireland) (see Fig. 3 therein). The observations used in this work were conducted predominantly over ocean and coastal regions (e.g. nearshore bases such as during the WISE campaign) and may have similar signatures to the coastal stations shown in . Neither model shows the pronounced seasonal pattern of CH2Br2 in the SH with elevated values in the southern hemispheric winter. This may be due to the emission scenario of , which was used in both models. The top–down emission estimates of the bromocarbons in the scenario are based on aircraft campaigns and some available observations in the marine boundary layer, most of which are from the NH. already identified the issue regarding the emission flux estimates in the SH as a consequence of missing aircraft observations in the SH (especially south of 40∘ S for all seasons). In addition, the emission inventory does not consider Br-VSLSs arising from sea ice regions, which are possible sources of Br-VSLSs according to, e.g., .

As noted before, CHBr3 has a much shorter atmospheric lifetime than CH2Br2. Consequently, the global tropospheric distributions from the observations of CHBr3 show larger variability and a less pronounced seasonality. used simulations to demonstrate that uniform background emissions of CHBr3 from the ocean result in a highly variable distribution in the atmosphere with larger values in regions of convergence or low wind speed and that the impact of localized elevated emissions on the distribution varies significantly from campaign to campaign. Thus, the interpretation of CHBr3 airborne observations is challenging. Nonetheless, the NH shows larger values in autumn and winter compared to spring and summer, a feature that is captured by both models, although the modelled wintertime maxima are more pronounced than in the observations and much less pronounced in autumn. Near-ground observations, however, may not be representative as they are largely from coastal areas. Based on the limited existing observations in the SH, the seasonality is not very pronounced with slightly smaller values in summer high latitudes and slightly larger near-ground values in autumn followed by winter. SH high-latitude ground-based NOAA/GML measurements presented in like the one from Cape Grim (Tasmania, Australia) and the Palmer Station (Antarctica), which are on a consistent scale as aircraft measurements, e.g. from PFP, also show a less pronounced seasonality. Furthermore, measurements of these stations reveal that models did not capture the observed seasonality for coastal stations see Fig. 3 in. Observations and model comparison from this study demonstrate a similar discrepancy as seen in at high latitudes of the SH with an underestimation in summer and autumn, where differences for tropospheric observations range between about 0.2 and 0.6 ppt, and overestimation mainly in winter high latitudes which reaches up to 0.8 ppt. Both models overestimate NH tropospheric values in winter and spring by up to 0.8 ppt except for near-ground values at high latitudes, which are much larger in the observational data (2–5 ppt difference to the observations above 80∘ N). In summer and autumn, the models underestimate high-latitude values (differences to observations of about 0.2 ppt and up to 0.9 ppt near the ground). Especially in the NH there is a larger frequency of observations over the northern Atlantic Ocean (see Fig. ). For example, the flights of the WISE campaigns took place predominantly from Shannon (Ireland), close to known coastline source regions of CHBr3, which may have a greater influence on the atmospheric distribution than photochemical decay e.g.. CHBr3 lower-stratospheric values are close to 0, as this substance has a shorter lifetime and is faster decomposed compared to CH2Br2. Model estimates of TOMCAT and CAM-Chem overall agree well within the LMS of both hemispheres.

Trace gases can enter the extratropical UTLS through several pathways. Beside the downward transport from the stratospheric overworld, there is a two-way exchange across the extratropical tropopause and an isentropic exchange, often in the vicinity of the subtropical jet e.g.and references therein. The amount of total bromine likely differs depending on how and where air enters the LMS and, consequently, the characteristics of the contributing input regions.

To investigate the distributions of the two major Br-VSLSs in the upper troposphere as a function of latitude, we binned the observational data according to latitude and to potential temperature relative to the local tropopause. Only data in the 10 K range below the local dynamical tropopause are included, to characterize the possible input region. For the models, only data in the 10 K range below the climatological PV-based tropopause are included. Data have been separated into low latitudes (0–30∘), midlatitudes (30–60∘), and high latitudes (60–90∘) for both hemispheres. Results for CH2Br2 and CHBr3 are listed in Tables  and , respectively, and the hemispheric winter CH2Br2 results are shown in Fig.  (graphical representations of the other seasons and seasonal cycle of each latitudinal band can be found in the Supplement). reported typical tropical tropopause mixing ratios compiled from different measurement campaigns with mixing ratios in the upper tropical tropopause layer (upper TTL) of 0.73 (0.43–0.94) ppt CH2Br2 and 0.28 (0.02–0.64) ppt CHBr3.

As shown in Table , low-latitude CH2Br2 values in both hemispheres during all seasons are well within the range reported by although with slightly higher mixing ratios. showed a clear tendency for an increase in tropopause mixing ratios with latitude in the NH, most pronounced in winter (see Fig. ). This is likely due to the increase in lifetimes with latitude, as photochemical breakdown becomes slower with higher latitudes especially during winter. Our extended data set confirms the findings of . In addition, we show that the SH upper-tropospheric distribution looks similar to that of the NH. Mixing ratios in NH winter and spring are slightly larger than in SH winter and spring, whereas mixing ratios in the respective hemispheric summer are close to each other across all latitudes. Larger differences are observed in the low and midlatitudes of NH and SH during autumn, with larger values in the NH than in the SH. However, the high latitudes of each autumn differ only to a small extent.

Both models qualitatively reproduce the larger CH2Br2 values in hemispheric winter and spring and smaller values in summer and autumn at high and midlatitudes (see Fig. S3 in the Supplement). However, both models overestimate low-latitude values for all seasons except for SH spring and NH winter. This overestimation in TOMCAT could already be seen in for the NH low latitudes for both winter and late summer to early autumn, as well as in where the TOMCAT model was only compared with HIPPO data. In general, CAM-Chem values are slightly smaller than TOMCAT values except in SH summer and autumn high latitudes. In NH winter, the models are close to observations (observations in between the model results), but both models overestimate NH spring values. However, both models reproduce the observed increase in mixing ratios with latitude in NH winter and spring, with CAM-Chem showing slightly better agreement. The models simulate SH winter and spring well in the midlatitudes and high latitudes, with TOMCAT somewhat closer to the winter observations, but none of the models capture the observed latitudinal variations in winter. The largest values in the models are found at high latitudes in spring for both hemispheres where observations show the largest values in hemispheric winter high latitudes (see Table  for absolute values). Summer and autumn of both hemispheres are in better agreement with CAM-Chem, as TOMCAT deviates more from the observations. TOMCAT generally simulates smaller mixing ratios in the SH and larger mixing ratios in NH summer and autumn midlatitudes and high latitudes.

For CHBr3 (see Table  and Figs. S2 and S4 in the Supplement), the upper-tropospheric distribution is much more variable and shows a less clear seasonality, especially in the SH. NH low-latitude values are slightly higher with a maximum in winter of 0.43 (0.33–0.62) ppt and maximum value in SH spring of 0.41 (0.36–0.43) ppt. NH values show an increase in mixing ratios with latitude, most pronounced in winter and spring. In NH summer, mixing ratios drop from low latitudes to midlatitudes and increase towards high latitudes again. Thus, the NH upper-tropospheric distribution of CHBr3 does not show the behaviour described in with an increase towards midlatitudes and a decrease at higher latitudes. considered a combined data set for summer and early autumn, whereas in this analysis summer and autumn are considered separately. As the latitudinal distribution differs substantially in summer and autumn (see Table ), the separate consideration of summer and autumn compared to the combined consideration as in can lead to differences in the interpretation of the observations. The behaviour of CHBr3, which can be seen in the NH summer, is also observed in the SH spring as well. All other seasons of the SH show an increase in mixing ratios with latitude but with the largest values in summer and autumn high latitudes with up to 0.57 (0.32–0.82) ppt in SH summer.

TOMCAT simulates an increase in mixing ratio with latitude for almost all seasons, with the largest values occurring in hemispheric winter. Only in hemispheric summer do values decrease from low latitudes to midlatitudes and then increase slightly in SH high latitudes and remain at the same level in the NH high latitudes. Mid- and high-latitude mixing ratios in the NH are highly overestimated in winter, spring, and to a smaller extent in autumn, a feature already observed in Fig. . In the SH, midlatitude mixing ratios are also overestimated in winter, spring, and autumn, which is also true for high-latitude spring and autumn. In contrast, the observed and modelled mixing ratios in the SH high latitudes in winter and summer show a different behaviour. In the observations, the high-latitude values are smallest in the winter and highest in the summer, which is a reversed behaviour of the model estimations (see Table 2 or Fig. S4). As the high-latitude observations are much more limited during these seasons (see Fig.  or Table S1 in the Supplement), we are careful about interpreting these differences and further observations are needed to verify these deviations from model results. CAM-Chem upper-tropospheric distributions are similar to TOMCAT distributions, although in general they show smaller values for all latitudes. Thus, CHBr3 mixing ratios from CAM-Chem are somewhat closer to the observation in NH winter, spring, and autumn midlatitudes to high latitudes and deviate slightly more in NH summer, compared to TOMCAT. In the SH, spring and autumn distributions from CAM-Chem are closer to observations compared to winter and summer distributions. Like for TOMCAT, the seasonal variation within the high latitudes show a different behaviour in comparison to the observations, with an even smaller value from CAM-Chem in SH summer high latitudes. Overall, the distribution of CHBr3 is highly variable, and both models simulate a similar latitudinal distribution, though with smaller values for CAM-Chem. The tug of war between rapid advective transport and local accumulation at the time of emission plays a decisive role. As already mentioned, showed that transport variations in the atmosphere itself produce a highly variable Br-VSLS distribution with elevated values not always reflecting strong localized sources.

The observational coverage of the upper troposphere and especially the lower stratosphere is best in spring and autumn of the respective hemisphere. From the altitude–latitude cross sections (Sect. 4.1, Figs.  and ), we already suspect larger differences in the LMS Br-VSLS distribution in hemispheric autumn than in hemispheric spring.

We thus took a closer look at the vertical profiles of CH2Br2 and CHBr3 in the midlatitudes of the SH and NH during these seasons. The observations from the different missions were seasonally combined and have been binned in 10 K intervals of potential temperature and potential temperature difference to the local tropopause. Only bins with at least five observations are considered. For the profile in the free troposphere, data were binned in potential temperature (Θ). Data binned in potential temperature difference to the local tropopause (ΔΘ) show larger variability in the free troposphere e.g.. ΔΘ coordinates are therefore not well suited for tropospheric data. As ΔΘ coordinates reduce the variability of the profile near the tropopause and within the lowermost stratosphere e.g., the profile continued from 10 K below the local tropopause into the stratosphere in ΔΘ coordinates. The two vertical coordinates were combined by aligning 0 K of ΔΘ with the median tropopause in Θ, observed during the measurements. Profiles in Θ coordinates may extend beyond the median tropopause even if the observations are declared as tropospheric ones. Tropopause Θ of these observations is much larger and corresponds to a higher tropopause, indicating that these observations may be subtropical in origin. The mixing ratios are averaged over equivalent latitude* of 40–60∘ of the respective hemisphere using box-and-whisker plots for the binned data (see Figs.  and ). Vertical gradients for spring and autumn profiles for both hemispheres from tropopause values up to 30 K above the local tropopause are summarized in Table .

Figure  shows hemispheric spring profiles of CH2Br2 and CHBr3. Upper-tropospheric values of CH2Br2 are very similar and close to 1 ppt in both hemispheres. The values at the tropopause are slightly larger (by 0.05–0.1 ppt) at the NH tropopause compared to the SH tropopause. The CH2Br2 profiles in the lowermost stratosphere of both hemispheres are very similar in their respective spring, beside an exceptional low value of 0.25 ppt at 90 K of ΔΘ in the NH. Vertical gradients of CHBr3 profiles are larger compared to CH2Br2, well in line with the much shorter lifetimes (see Table ). Upper-tropospheric mixing ratios are slightly larger in SH midlatitudes (by roughly 0.16 ppt). However note that NH tropospheric values present a larger variability. In contrast, tropopause values are slightly larger in the NH than in the SH. Mixing ratios drop to values close to 0 in both hemispheres at about 30–40 K of Θ above the tropopause.

Figure  shows hemispheric autumn profiles of CH2Br2 and CHBr3 with less similarity between SH and NH profiles than in hemispheric spring for both compounds. Upper-tropospheric values of CH2Br2 are slightly larger in the NH with a difference of up to 0.16 ppt to the SH. Tropopause values for hemispheric autumn (Fig. 7a and b) are slightly smaller compared to hemispheric spring values (Fig. 6a and b), but hemispheric differences in spring and autumn are comparable. Differences between NH and SH autumn become larger on the lowest levels above the dynamical tropopause; i.e. in the ExTL. CH2Br2 shows a larger vertical gradient up to 30 K of ΔΘ in SH autumn than in NH autumn (see Table ) reaching smallest value of 0.45 ppt between 20 and 30 K of ΔΘ. With a comparable distance to the dynamic tropopause, the value in the NH is 0.64 ppt. Between 40 and 70 K of ΔΘ, SH values range from 0.53 to 0.7 ppt associated with a large variability.

Figure  again shows the SH autumn vertical profile of CH2Br2, but two more profiles were included in ΔΘ coordinates with one profile excluding TOGA observations and one profile excluding WAS and PFP observations. The profile including only TOGA from ATom is much closer to the profile using all observations, whereas the profile including only WAS and PFP observations from ATom shows a much steeper gradient across the tropopause and much smaller values. The larger number of TOGA observations shifts the median towards larger values. The profile without TOGA measurements would be in line with the assumption of a strong transport barrier, e.g. the subtropical jet and exchange between stratosphere and troposphere only at the edges of the jets e.g. Fig. 1 in, while the profile without WAS and PFP shows a less strong but still larger transport barrier than in the NH autumn profile (Fig. b). The NH profile also shows a much smaller gradient compared to NH spring and less variability, which may indicate a well-mixed LMS. This is also well in line with finding by, e.g., , who showed a flushing of the lowermost stratosphere in summer and autumn.

Hemispherical differences can also be inferred from CHBr3 vertical profiles. Larger differences can be seen in the upper troposphere with nearly double the amount in the NH compared to the SH. In contrast, hemispheric differences at the tropopause are small (close to 0.5 ppt for both hemispheres). Regarding the LMS and especially the ExTL, the vertical gradient is much larger in the SH than in the NH. This implies that there is a stronger transport barrier in the SH when looking at CHBr3. SH values drop to almost 0 in the first 20 K above the dynamical tropopause but to around 0.22 ppt at a comparable distance to the NH autumn tropopause. Furthermore, NH lowermost-stratospheric values are steadily decreasing and are above 0.02 ppt even at 90 K of ΔΘ.

We further compared the midlatitude profiles of CH2Br2 and CHBr3 with model results of TOMCAT and CAM-Chem in Figs.  and . The comparisons are shown as a function of ΔΘ. Since there is no tropopause information for the TOMCAT model, we derived ΔΘ as the difference between the climatological tropopause potential temperature and model potential temperature. For consistency, the climatological tropopause was used for CAM-Chem as well. Additionally, since equivalent latitude information is not available for the models, latitude was used instead. The mean absolute percentage differences (MAPDs) reported below are tabled in the Supplement.

For spring CH2Br2 (Fig. a and b), TOMCAT profiles overestimate the observations on average by about 0.12 ppt in the NH and 0.08 ppt in the SH, with a corresponding MAPD of about 26 % (NH) and 18 % (SH). The CAM-Chem profiles are closer to the observations. Particularly the NH profile within the lowest 20 K above the dynamical tropopause is close to the observations but deviates towards higher altitudes. Average differences in the NH are 0.06 ppt and an MAPD of 17 %, whereas in the SH, differences are on average 0.03 ppt with an MAPD of 9 %. For CHBr3 (Fig. c and d), the comparison is only shown up to 40 K of ΔΘ, since this substance was almost completely depleted above. TOMCAT seems to overestimate NH CHBr3 by on average 0.1 ppt, corresponding to an MAPD of 69 %. Differences are much smaller in the SH with 0.02 ppt and an MAPD of 21 %. CAM-Chem estimated the lowermost stratosphere well in the NH with an average difference of 0.02 ppt and an MAPD of 15 %. SH differences in CAM-Chem are on average 0.04 ppt with an MAPD of 26 %, thus slightly larger than TOMCAT differences. Both models estimated a nearly complete depletion of CHBr3 in the SH lower stratosphere (above 40 K of ΔΘ). However, observations show a slight offset to the model estimations with values of 0.02–0.04 ppt up to 90 K of ΔΘ (Fig. c).

Hemispheric autumn midlatitude profiles of CH2Br2 and CHBr3 are displayed in Fig. . In NH autumn, the observed profiles and those of the models agree well. In the SH, however, the profiles diverge further as neither of the models reproduce the presumably stronger barrier at the tropopause described in Fig. . TOMCAT overestimates NH CH2Br2 up to 90 K of ΔΘ on average by about 0.1 ppt (MAPD of 20 %). The difference between models and observations becomes much larger in SH autumn. In addition, the observational profile exhibits a high degree of variability, and the scatter of the individual bins increases due to the differences between the ATom instruments, as mentioned earlier. Although models are close to observations above 40 K of ΔΘ, average differences between 0 and 40 K of ΔΘ are 0.12 ppt (MAPD of 25 %). CAM-Chem estimates are closer to the observations in the NH and differ on average by about 0.02 ppt corresponding to an MAPD of 6 %. However, CAM-Chem showed a similar profile to TOMCAT in the SH, although slightly smaller values in general and thus a smaller average difference between 0 and 40 K of ΔΘ of 0.06 ppt and an MAPD of 15 %. Compared to NH spring, TOMCAT shows a CHBr3 profile which is much closer to the observations in NH autumn. The average difference below 40 K of ΔΘ is 0.06 ppt with a corresponding MAPD of 22 %. SH observations are limited and thus a comparison is only possible up to 20 K of ΔΘ. Although observations and model results of TOMCAT are close near the dynamical tropopause, the difference increases rapidly to 0.13 ppt between 10 and 20 K of ΔΘ. As the CAM-Chem profile is smaller in absolute values, the largest differences are around 0.08 ppt. Both models fail to capture the steeper gradient across the dynamical tropopause during SH autumn.