Water vapour and ozone are important for the thermal and radiative balance of the upper troposphere (UT) and lowermost stratosphere (LMS). Both species are modulated by transport processes. Chemical and microphysical processes affect them differently. Thus, representing the different processes and their interactions is a challenging task for dynamical cores, chemical modules and microphysical parameterisations of state-of-the-art atmospheric model components. To test and improve the models, high-resolution measurements of the UT–LMS are required. Here, we use measurements taken in a flight of the GLORIA (Gimballed Limb Observer for Radiance Imaging of the Atmosphere) instrument on HALO (High Altitude and LOng Range Research Aircraft). The German research aircraft HALO performed a research flight on 26 February 2016 that covered deeply subsided air masses of the aged 2015/16 Arctic vortex, high-latitude LMS air masses, a highly textured region affected by troposphere-to-stratosphere exchange and high-altitude cirrus clouds. Therefore, it provides a challenging multifaceted case study for comparing GLORIA observations with state-of-the-art atmospheric model simulations in a complex UT–LMS region at a late stage of the Arctic winter 2015/16.
Using GLORIA observations in this manifold scenario, we test the ability of
the numerical weather prediction (NWP) model ICON (ICOsahedral
Nonhydrostatic) with the extension ART (Aerosols and Reactive Trace gases)
and the chemistry–climate model (CCM) EMAC (ECHAM5/MESSy Atmospheric Chemistry – fifth-generation European Centre Hamburg general circulation model/Modular Earth Submodel System) to model the UT–LMS composition of water vapour (H
Trace gas composition, in particular the vertical distributions of greenhouse gases, and clouds play an important role in the thermal and radiative budget of the upper troposphere–lowermost stratosphere (UT–LMS) (e.g. Riese et al., 2012; Hartmann et al., 2013). Stratospheric and, particularly, lowermost-stratospheric water vapour has been identified to be an important driver in decadal global surface climate change (e.g. de Forster and Shine, 2002; Solomon et al., 2010). Also, changes in stratospheric ozone are well known to affect temperature trends and radiative forcing (e.g. de Forster and Shine, 1997). In the lower stratosphere, ozone depletion is a major contributor to its negative temperature trend. There is also a significant spread among modelled trends when ozone and other greenhouse gas abundances are perturbed. Explanations for such differences include the different responses of individual radiation schemes and different sensitivities in the dynamical forcing in the models to changes in trace gases (e.g. Shine et al., 2003). Lowermost-stratospheric water vapour distributions show hemispheric differences, thus requiring knowledge of hemispheric and latitudinal distributions and change for accurate climate projections (e.g. Kelly et al., 1991; Rosenlof et al., 1997; Pan et al., 1997).
The LMS is the lowest layer of the stratosphere situated between the local tropopause and the 380 K isentropic level (e.g. Werner et al., 2010). In the winter hemisphere, its composition is mainly affected by air mass contributions from the polar winter vortex, the mid-latitude stratosphere and the troposphere. While air masses in the polar winter vortex are mostly isolated from the surrounding stratosphere, LMS air masses at the bottom of the polar vortex can be affected significantly by interactions with air masses from lower latitudes (e.g. Krause et al., 2018).
Rossby waves are undulations of the eastward-directed upper-tropospheric flow in the midlatitudes and are accompanied by step-like changes in the height of the dynamical tropopause (e.g. Wirth et al., 2018). Rossby wave breaking events can be identified as overturning patterns in Ertel's potential vorticity (PV) and contribute to exchange of upper-tropospheric and lower-stratospheric air masses (e.g. Gabriel and Peters, 2008; Jing et al., 2018).
Exchange processes including quasi-isentropic and cross-isentropic exchange occur often in the vicinity of jet streams (e.g. Holton et al., 1995; Gettelman et al., 2011). They can be accompanied by different kinds of tropopause folds and modulate the trace gas composition of the UT–LMS. Irreversible fluxes between the UT and the LMS can occur in either direction – from stratosphere to troposphere and from troposphere to stratosphere. Generally, the dominating flux in the extratropics is directed towards the troposphere. Such exchange processes and their effects have been investigated by numerous field observations (e.g. Ray et al., 1999; Hoor et al., 2002, 2005; Bönisch et al., 2009; Krause et al., 2018) and by many theoretical and modelling studies (e.g. Meloen et al., 2003; Stohl et al., 2003, and references therein).
Cirrus clouds are one of the least-understood factors modulating climate
change and affecting the composition of the UT–LMS (e.g. Schiller et al.,
2008; Barahona and Nenes, 2009). Cirrus clouds absorb upwelling infrared
light and reflect sunlight back to space and thereby affect the radiative
budget and thus the thermal structure of the tropopause region.
Sedimentation of cirrus cloud ice particles redistributes water vertically
and changes the water vapour profile. Furthermore, the ice particles are
capable of trapping nitric acid and other trace gases (e.g. Popp et al.,
2004; Voigt et al., 2006; Krämer et al., 2008; Kärcher et al.,
2009). Moreover, vertical distributions of H
Nowadays, numerical weather prediction and chemistry–climate models (NWPs and CCMs) are capable of resolving the UT–LMS, mesoscale dynamics and cloud processes in part explicitly and in part by using parameterisations ranging from low to high complexity. Examples of such models include ICON (ICOsahedral Nonhydrostatic; see Zängl et al., 2015) with the extension ART (Aerosols and Reactive Trace gases; see Rieger et al., 2015 and Schröter et al., 2018) and EMAC (ECHAM5/MESSy Atmospheric Chemistry – fifth-generation European Centre Hamburg general circulation model/Modular Earth Submodel System; see Jöckel et al., 2006, 2010, 2016, and Roeckner et al., 2006). However, accurate simulations of UT–LMS composition, dynamics and cirrus clouds (and their interactions) remain a challenge and are important building blocks for reliable weather forecasting and climate projections. In particular, LMS water vapour is known to be affected by significant systematic errors in model simulations (e.g. Stenke et al., 2008).
The exceptionally cold Arctic winter 2015/16 was characterised by a stable polar vortex and low temperatures in the UT–LMS region (Matthias et al., 2016). While the winter was the coldest on record from December to early February, complex dynamical processes and a major final stratospheric warming in early March ended the cold phase and resulted in a vortex split in mid-March (Manney and Lawrence, 2016). In the same winter, airborne observations in the framework of the combined POLSTRACC (POLar STRAtosphere in a Changing Climate), GW-LCYCLE (Gravity Wave Life Cycle Experiment) II and SALSA (Seasonality of Air mass transport and origin in the Lowermost Stratosphere using the HALO Aircraft) (PGS) field campaign probed the Arctic UT–LMS region in the period from December 2015 to March 2016 (Oelhaf et al., 2019). During PGS, the GLORIA (Gimballed Limb Observer for Radiance Imaging of the Atmosphere) instrument (Friedl-Vallon et al., 2014; Riese et al., 2014) was deployed on board the German HALO (High Altitude and LOng Range Research Aircraft). From the GLORIA limb-imaging observations, vertical distributions of temperature, trace gases and clouds are derived and allow detailed model comparisons (e.g. Khosrawi et al., 2017; Braun et al., 2019; Johansson et al., 2019).
During the research flight on 26 February 2016 (PGS 14), GLORIA probed
subsided LMS air masses of the aged 2015/16 polar vortex at high latitudes,
a highly textured region affected by troposphere–stratosphere exchange, and
high-altitude cirrus clouds across a long transect spanning from Scandinavia
over Greenland to Canada. Here, we use the GLORIA observations during this
flight to test the capabilities of EMAC and ICON-ART of modelling mesoscale
H
In Sect. 2, we introduce our observations, models and diagnostics. An overview of the meteorological situation and the GLORIA observations during PGS 14 is provided in Sect. 3. In Sect. 4, the two-dimensional vertical cross sections of modelled cloud and trace gas distributions are compared with the GLORIA observations, discrepancies are diagnosed and investigated, and sensitivity experiments with the models are presented. We furthermore investigate the evolution of narrow moist filaments observed by GLORIA in the LMS with the aid of ICON-ART. The results are summarised and discussed in Sect. 5.
In the following, the characteristics of the GLORIA observations, the model set-ups used and the applied diagnostics are introduced. An overview of the cloud and trace gas products used is provided in Tables 1 and 2.
The GLORIA data used here were measured during the HALO flight PGS 14 on 26
February 2016. PGS 14 started in Kiruna, northern Sweden, and covered the
Arctic Sea, Greenland and eastern Canada (Fig. 1b). GLORIA is a passive
infrared limb-imaging spectrometer deployed on board high-altitude aircraft
(Friedl-Vallon et al., 2014; Riese et al., 2014). GLORIA uses 128 vertical
Optical information on vertical cloud coverage is obtained directly from the
calibrated spectra by using the cloud index method (Spang et al., 2004). The
cloud index uses the colour ratio between the spectral microwindows from
788.20 to 796.25 and 832.30 to 834.40 cm
The state-of-the-art global meteorological forecast system ICON (Zängl et al., 2015) has been operational at the German Weather Service (Deutscher Wetterdienst, DWD) since 2015. ICON was developed by the DWD in cooperation with the Max Planck Institute for Meteorology, Hamburg. ICON uses a triangular grid, which is well suited for modern computer architectures. Further, it allows efficient scaling of the dynamical core, avoids meridional grid convergence and singularities at the poles, improves mass conservation, and allows efficient local grid refinement with two-way interaction (nesting). In the vertical domain, a hybrid height coordinate is used (Leuenberger et al., 2010) that continuously transforms from local topography-following levels to constant height levels at 16 km and above.
The Aerosols and Reactive Trace gases module ART was developed at the Karlsruhe Institute of Technology (KIT). It simulates chemical processes and aerosols and couples trace gas concentrations and aerosols at each model time step to other relevant processes (Rieger et al., 2015; Schröter et al., 2018). The ICON transport scheme redistributes the tracers, and clouds and radiation properties are coupled to the meteorological state. ART is capable of simulating chemical and photo-chemical production and loss of reactive trace gases and can be used with defined emission scenarios (Weimer et al., 2017).
For the PGS campaign, a dedicated ICON-ART simulation was performed for the
entire polar winter 2015/16 using an R2B6 (
For the POLSTRACC winter, a global ICON-ART simulation
with a global R2B6 grid was carried out (red). In the area of the flights, a
nest with an R2B7 grid with
Other than the meteorological variables, tracers, such as the ozone tracers,
are simulated continuously in a free-running mode after initialisation at
the beginning of the winter using a previous EMAC simulation (Schröter
et al., 2018) and are not reinitialised regularly at 00:00 UTC. The simulation
of polar stratospheric ozone loss in the simulated “O
For qualitative comparisons with clouds observed by GLORIA, the sum of
specific cloud ice content (
Data sets and cloud parameters (cirrus and ice clouds).
Data sets, trace gas products and sensitivity simulations.
The ECHAM5/MESSy Atmospheric Chemistry (EMAC) model is a numerical chemistry
and climate simulation system that includes submodels describing
tropospheric and middle-atmospheric processes and their interaction with
oceans, land and human influences (Jöckel et al., 2010). It uses the
second version of the Modular Earth Submodel System (MESSy2) to link
multi-institutional computer codes. The core atmospheric model is the fifth-generation European Centre Hamburg general circulation model (ECHAM5;
Roeckner et al., 2006). In this study we used EMAC (ECHAM5 version 5.3.02,
MESSy version 2.52; see Jöckel et al., 2010) with T42L90MA and T106L90MA
resolution, i.e. with a spherical truncation of T42 (corresponding to a
quadratic Gaussian grid of 2.8
The EMAC standard and sensitivity simulations employed Eulerian
grids with 106 (red) and 42 (blue) spectral coefficients. The T106 (T42)
grid corresponds to a horizontal resolution of 125 km (310 km) at the
Equator. Due to the grid convergence, the zonal grid spacing is reduced
towards the poles and amounts to
The applied model set-up includes a comprehensive chemistry scheme with gas-phase reactions and heterogeneous reactions on polar stratospheric clouds (PSCs) and comprises about 35 submodels, including the chemistry submodel MECCA (Sander et al., 2011); the photolysis submodel JVAL (Sander et al., 2014); the submodel MSBM, mainly responsible for the simulation of PSCs (Kirner et al., 2011); the submodel CLOUD, based on the ECHAM5 cloud scheme, simulating large-scale clouds (Roeckner et al., 2006); the submodel CONVECT, calculating the convection and convective clouds (Tost et al., 2006b); and the submodel SCAV, responsible for scavenging and wet deposition of trace gases and aerosols (Tost et al., 2006a).
We performed three different simulations from 1 July 2015 to 1 April 2016
(initialised with an older EMAC simulation, which was started in 1994 and
perpetuated until recent years), thus including the Arctic winter 2015/16 and
the PGS campaign. In the first simulation (our “standard” simulation), we
use the horizontal resolution of T106 (EMAC-STD). Additionally, we
performed two sensitivity simulations: first we reduced the horizontal
resolution to T42 (EMAC-T42). In the second, we switched off the scavenging
processes on ice particles, using again the T106 resolution (EMAC-NOSCAV).
For comparisons
with clouds observed by GLORIA, the combination of EMAC large-scale cloud
snow and ice content (iwc) and convective cloud snow and ice content
(cv_iwc) is used (see Table 1). With respect to trace gases,
the following EMAC variables are used: water vapour (H
The vertical profiles of clouds and trace gases are combined into time–height cross sections of these parameters along the HALO flight tracks. For direct comparisons of synoptic and mesoscale patterns with the models, the ICON-ART and EMAC fields of the respective parameters are interpolated to the tangent point geolocations of the GLORIA observations (Fig. 1) to yield the corresponding model cross sections. In the vertical cross sections of the GLORIA data products, PV contours from the corresponding ECMWF reanalysis are superimposed to indicate the dynamical tropopause. For the model cross sections, PV is interpolated from the respective model output.
To quantify biases in the modelled trace gas distributions, the GLORIA and
the interpolated model data of the variable under consideration are
correlated against each other. In this manner, discrepancies between model
simulations and observations can be identified as systematic deviations of
data point populations from the respective
The vertical resolution of the GLORIA data used here is on the order of 500 m, depending on altitude and parameter (see Johansson et al., 2018a), and therefore comparable with the vertical resolution of the simulations by both models in the tropopause region. Therefore, the use of 1-D averaging kernels in the vertical domain, as are often used in the context of vertical profiles retrieved from satellite limb observations (e.g. Microwave Limb Sounder, MLS) that are characterised by notably coarser vertical resolution, is not expected to improve the comparison significantly. The absence of relevant overall systematic biases in the GLORIA data used here is furthermore confirmed by in situ comparisons (see Johansson et al., 2018a).
Due to the limb viewing geometry, strong horizontal gradients along the line of sight of GLORIA (i.e. towards the right-hand side of the flight track) can affect direct comparisons of vertical cross sections of atmospheric parameters derived from the GLORIA observations and interpolated from the models at the tangent points. This effect can be taken into account by interpolating the model data with the help of 2-D averaging kernels (Ungermann et al., 2011, their Sect. 3.2). As discussed by Woiwode et al. (2018) in a case study where the mesoscale fine structure of a tropopause fold was investigated, the application of 2-D averaging kernels improves the model comparison only moderately if the observations are aligned such that horizontal gradients in the trace gas fields along the line of sight are small (see their Appendix A).
Aided by meteorological forecasts, the flight analysed here was planned so that the GLORIA observations were mostly aligned in such a way. This can be seen by comparing Fig. 1b with Fig. 4b, for example during the backward leg to Kiruna, when the GLORIA limb views were aligned along the direction of moist filaments above Greenland. Therefore, the viewing geometry allowed us to resolve the fine structures of the narrow filaments discussed in Sect. 4.3 remarkably well. Due to the suitable alignment of the GLORIA observations during the discussed flight and since the application of 2-D averaging kernels is computationally demanding (particularly in case of the GLORIA high spectral resolution chemistry mode observations that employ a large number of spectral sampling points), 2-D averaging kernels are not applied here. Therefore, local discrepancies between the GLORIA and model cross sections due to remaining effects by horizontal gradients along the line of sight cannot be excluded.
Meteorological conditions in the tropopause region and at sea
level on 25 February 2016
However, when the complete ensemble of GLORIA and model data points is analysed, such remaining effects by horizontal gradients are expected to cancel out on average due to the large number of data points. Therefore, we consider the estimation of model biases in Sect. 4.4 to be robust.
Due to low planetary wave activity the Arctic winter 2015/16 was
extraordinarily cold (relative to preceding decades), and a strong polar
vortex formed during November and December 2015 (Matthias et al., 2016).
Cold conditions prevailed until February 2016. Then, three minor
stratospheric warmings led to slightly warmer conditions in the polar
vortex, but temperatures remained below the nitric acid trihydrate (NAT) PSC
existence temperature (
PGS 14 was performed on 26 February 2016 from Kiruna, northern Sweden. Take-off of the HALO aircraft was at 11:19 UTC and landing time was at 20:59 UTC (flight duration of 9 h 40 min). The HALO flight track (anti-clockwise) and the tangent points of the GLORIA limb observations are shown in Fig. 1b. After take-off, HALO headed westward (GLORIA pointing northward), crossed the Atlantic and Greenland, and continued its flight towards Canada. Then, at waypoint A, it turned southward (GLORIA pointing westward). Finally, after waypoint B, HALO turned back eastward and headed back towards Scandinavia (GLORIA pointing southward).
Figure 4 shows the meteorological situation on the day before the flight at
12:00 UTC (left column) and for the flight day at 18:00 UTC (right column), i.e.
during the eastward flight leg back to Kiruna. The colour-coded contour
plots in the upper row show ICON-ART
The surface weather conditions are shown in the lower row of Fig. 4. On 25 February 2016, a well-defined low-pressure system is located above
Scandinavia, and patchy weak high-pressure systems are found around central
Greenland and Canada. A strong Azores high is located in the Atlantic Ocean
together with a compact Icelandic low located at the southern tip of
Greenland, accompanied by a notably positive North Atlantic Oscillation
(NAO) index of
Overall, at 10 km the air masses observed by GLORIA on 26 February 2016 comprise (i) the high-latitude LMS including patchy filaments, (ii) deeply subsided polar vortex air masses above Canada, (iii) upper-tropospheric air masses above southern Greenland, (iv) moist air filaments above Greenland and associated with the occluded front of the Icelandic low, and (v) again high-latitude LMS air masses. Therefore, the GLORIA observations provide a unique opportunity to test the capability of ICON-ART and EMAC in simulating the Arctic winter UT–LMS region.
The vertical cross section of the GLORIA cloud index of the entire flight is
shown in Fig. 5a. Cloud index (CI) values close to 1 are indicative of
optically thick conditions, i.e. in the presence of clouds, whereas CI
values approaching 4 and higher can be considered to be cloud-free
conditions in spaceborne limb-sounding observations (Spang et al., 2004). In
the case of airborne limb observations, CI values of 2 to 4 have been found
to be suitable to differentiate between cloud-affected and cloud-free conditions
in previous studies (Johansson et al., 2018a, and references therein). In the
case presented here, a cloud index of
Qualitative comparison of clouds along the flight track observed by
GLORIA and cloud masks generated from ICON-ART and EMAC.
In the following, we compare GLORIA cloud index values with cloud masks generated from the models in a qualitative way. The GLORIA cloud index is an optical quantity, while the model cloud masks are generated from the respective model outputs for condensed water in the solid state (see Table 1). Liquid water is not considered since the temperatures in the focus region are well below the frost point, and there was no significant contribution of liquid water to the cloud masks used. A quantitative comparison (e.g. conversion of modelled cloud properties into spectral radiances and considering effects related to line of sight) is beyond the scope of our study, which focuses on the ability of the models to reproduce the smaller-scale structures.
We have set the threshold for the ICON-ART and EMAC model cloud mask at
10
The ICON-ART cloud mask represents the sum of cloud ice content (
In the global and nested ICON-ART domains (Fig. 5b and d), three of the four major cloud systems seen in the GLORIA observations can be identified, with differences in the vertical and horizontal extent. The results of the global and nested domains show only small differences, which are attributed to the simulations on the different grids and the interpolation from these grids with different widths.
The observed cloud system around 14:00 to 15:00 UTC below 10 km altitude is missing in both shown ICON-ART representations. Modelled cloud systems below approx. 10 km around 12:00 to 13:00 UTC, 16:30 to 17:30 UTC and 20:00 UTC agree well with GLORIA in the horizontal domain. Discrepancies in the large cloud system around 20:00 UTC below 6 km can be explained by the fact that no robust information on vertical cloud structure can be derived from GLORIA if optically dense cloud layers are located above. In such cases, lower limb views can be optically saturated, and low cloud index values may result, although cloud-free conditions are present below. The same effect might explain differences between the observed and modelled cloud system between 12:00 and 13:00 UTC. We explain the fact that the vertically extended cloud system detected by GLORIA around 14:00 to 15:00 UTC is not reproduced by the ICON-ART simulation in both domains (global and nested) by a temporal mismatch in the simulated cloud systems (see Appendix A). Furthermore, the discrepancies might be explained partly by line-of-sight-related effects since GLORIA accumulates light along extended limb views, while the model is interpolated at a certain geolocation. For the observed cloud systems at lower altitudes between 17:30 and 19:30 UTC, only weak indications are found in the ICON-ART simulation. Further high cloud systems prior to 12:00 UTC are barely reproduced in the ICON-ART simulation, and a simulated cloud at 16:00 UTC below 6 km is not confirmed by GLORIA.
The corresponding cloud mask of the EMAC standard simulation (EMAC-STD) with the T106L90MA resolution was generated by using the sum of the large-scale cloud snow and ice content (iwc) and the convective cloud snow and ice content (cv_iwc) (see Table 1). As mentioned earlier the EMAC simulation uses a continuously nudged meteorology (see Sect. 2.3); however, the cloud variables are not nudged. As can be seen in Fig. 5c, the EMAC-STD reproduces the cloud patterns observed by GLORIA well. All of the observed cloud systems can be found in the cross section along the flight path generated from the EMAC simulation. Especially, the observed cloud system between 14:00 and 15:00 UTC, which is not reproduced by ICON-ART, is reproduced by EMAC, but with a different morphology and slightly displaced horizontally and vertically. Also, the lower clouds observed between 17:30 and 19:30 UTC are reproduced well by the EMAC simulation. As in the case of ICON-ART, a simulated low cloud system at 16:00 UTC is not confirmed by GLORIA.
In the EMAC simulation the modelled horizontal and vertical extents are mostly larger when compared to ICON-ART (e.g. prior to 12:00 UTC and 16:30 to 19:30 UTC). The lower model resolution and lower time resolution of the output (1 h for EMAC versus 0.25 h for ICON-ART) could be one possible explanation, making a positive cloud detection more likely (concerning the spatial coverage). Furthermore, the lower grid spacing is more comparable to the horizontal extensions of the GLORIA limb views, which results in a more consistent comparison in certain cases when cloud systems are located along the line of sight. The high cloud system prior to 12:00 UTC matches the GLORIA cloud index better than in the case of ICON-ART, while the cloud system in EMAC from 12:00 to 13:00 UTC appears higher than in the GLORIA and ICON-ART data, even exceeding the 2 and 4 PVU isoline and reaching the GLORIA flight altitude. The clouds from 16:30 to 17:30 UTC and at 20:00 UTC also reach higher in the atmosphere in the EMAC cross section compared to the GLORIA and ICON-ART data and again notably higher than the respective local dynamical tropopause. In these cases the ICON-ART cloud mask agrees better with the GLORIA observations.
Another proxy for the characterisation of detectable cloud systems in the model is looking at the cirrus cloud ice particle sedimentation events, which include the processes of nucleation, sedimentation and subsequent evaporation of cirrus cloud ice particles. As a consequence, local irreversible dehydration is found when ice particle growth removed water from the gas phase, and hydration is found at lower altitudes, where the particles sublimate.
This is done in the following in the case of ICON-ART by using a passive
water vapour tracer forecast in the constrained forecast mode as a
reference. In addition, this analysis sheds light on the degree to which
cirrus cloud ice particle sedimentation affects the modelled water vapour in
the UT–LMS (cf. Sect. 4.2). The passive water vapour tracer does not account
for cloud microphysics and therefore no nucleation, sedimentation and
evaporation of hydrometeors. Residuals between the ICON-ART specific
humidity forecast (see Sect. 4.2) and the passive reference tracer show
where microphysical processes altered UT–LMS humidity within the time frame
of the forecast (i.e. the forecast lead time between
Figure 6 shows the residual, i.e. the difference between ICON-ART (nested
domain) specific humidity and the passive tracer without cloud microphysics.
Negative residuals indicate regions which are depleted in water vapour due
to cloud processes. Positive residuals show regions enriched in gas-phase
water vapour due to sublimation of ice and snow particles. Negative and positive
residual patterns clearly prove the generation and transformation of
hydrometeors in the UT–LMS during the entire flight. Before the waypoint A,
a strong pattern with residuals exceeding
Modelled short-term changes in specific humidity due to cloud
processes. Residuals between nested ICON-ART domain of specific humidity and
corresponding H
The comparison of Fig. 6 with Fig. 5a shows that this idealised ICON-ART diagnostic is a good proxy for the simulation of clouds in the model and does not require a threshold approach (as discussed above). However, it is an integrated quantity showing the history of “cloud events” on the respective day, whereas the cloud masks show “snapshots” of simulated hydrometeors at the geolocations and time of the measurement. At a closer look, all of the observed cloud systems coincide qualitatively with a corresponding cirrus cloud ice particle sedimentation pattern at the respective geolocations in the ICON-ART data. This means that there is evidence for the existence of all observed cloud systems in the ICON-ART simulation. However, as in the case of the ICON-ART cloud mask prior to 12:00 UTC, only weak indications of cloud systems are found here.
In particular, between 14:00 and 15:00 UTC, where a cloud system detected by GLORIA
is not reproduced by the cloud mask of ICON-ART (as described above; cf.
Fig. 5a, b, d), barely resolved weak negative residuals reaching up to
about
There is also evidence in the ICON-ART data for the lower cloud system observed between 17:30 and 19:30 UTC (compare Fig. 6 with Fig. 5a). Even though this cloud system is underestimated in the simulation (see Fig. 5b, d), Fig. 6 suggests that it has been present at these locations at some time prior to the measurement in the simulation on the day of the flight.
The narrow cloud band at waypoint A, detected by GLORIA around 8 km and also evident in the EMAC cross section (see Fig. 5c), is not visible in the ICON-ART cross section (compare Fig. 5a and c with b and d). However, again a strong signal of vertical redistribution of water vapour is visible in Fig. 6 at this geolocation, which again hints at the presence of this cloud system in the ICON-ART simulation at some time prior to the measurement. Thus, uncertainties in the timing of the ICON-ART forecast might partly explain the discrepancies between GLORIA and ICON-ART here in addition to the other reasons discussed above.
In Appendix A we further investigate this issue by sampling the models at the respective GLORIA geolocations with a negative time offset to shed light on the history and development of the cloud systems in the models on the day of the flight and to prove that seemingly “missing clouds” in the ICON-ART data based on the cloud mask can be identified in the simulations just a few hours prior to the measurements.
Overall, the simulated cirrus cloud ice particle sedimentation patterns in
Fig. 6 are consistent with the observed and modelled cloud systems in Fig. 5
and clearly show that modelled water vapour distributions in the UT–LMS are
significantly modulated by more than
Note that line-of-sight-related effects are capable of particularly strong influences on the comparison with respect to clouds. If, for instance, clouds were situated in front or behind the tangent point along the line of sight, this comparison would lead to a discrepancy between the model results and the measurements. Especially, complex small-scale cloud structures with strong optical gradients (transparent or opaque) can differ in coverage and orientation when compared to the trace gas fields. Therefore, we consider the comparison of GLORIA cloud detection to the simulated clouds to be more difficult, in particular for small clouds or edges of clouds. Despite these limitations of the comparison, we mostly found good agreement between GLORIA and the models.
In the following, we compare observations of water vapour, ozone and nitric
acid with the respective simulated trace gases by ICON-ART and EMAC. For the
former only water vapour, i.e.
Figure 7a–c show the water vapour, ozone and nitric acid distributions
observed by GLORIA along the flight track. When compared with the cloud
index plot (Fig. 5a), gaps in the retrieved trace gas distributions are
explained by the fact that the presence of dense clouds precludes trace gas
retrievals in the affected regions. Cloud filtering is applied here prior to
the trace gas retrieval. Before waypoint A, moist tropospheric air masses
extend to the dynamical tropopause, which is located mostly around 10 km in
Fig. 7. Some moist “patches” are also found in the LMS here. In contrast,
dry stratospheric air masses reaching down to
Observed and modelled trace gas distributions. GLORIA observations
of water vapour, ozone and nitric acid
The ozone distribution (Fig. 7b) shows a converse pattern compared to water
vapour. At tropospheric altitudes, low ozone mixing ratios are found, while
ozone mixing ratios above the tropopause increase with altitude. Also, in
the ozone distribution, the deeply subsided polar vortex remnant from
waypoint A to slightly after waypoint B can be clearly identified by high
ozone mixing ratios reaching down towards
Furthermore, the nitric acid distribution shows a local maximum at and below
flight altitude from
As shown in Fig. 7d, the overall distribution and mesoscale structures in
the ICON-ART specific humidity forecast on the global R2B6 grid agree well
with water vapour detected by GLORIA. Recall that
Keeping in mind that water vapour is simulated by EMAC continuously (i.e. no
reinitialisation at 00:00 UTC and not nudged), the EMAC-STD simulation also
reproduces the observed water vapour distribution well (Fig. 7f). Naturally,
fewer details are found in the EMAC simulation due to the lower horizontal
resolution. The subsided air mass from A to slightly after B is reproduced
by EMAC. However, moister air masses with water vapour
The continuous ICON-ART ozone simulation (i.e. no reinitialisation at 00:00 UTC) on the global R2B6 grid also matches the mesoscale patterns seen in the GLORIA observations (Fig. 7e), however with systematically lower volume mixing ratios. Again, the deeply subsided air masses from waypoint A to slightly after waypoint B can be clearly identified by higher ozone mixing ratios reaching down to lower altitudes. Similar filaments and structures as seen in the GLORIA observation between 17:30 and 19:00 UTC are identified, however with fewer details and fine structures. The EMAC ozone distribution (Fig. 7g) matches the GLORIA observations well within the limitations of the model resolution, as already discussed by Johansson et al. (2019). Here, absolute mixing ratios agree quite well with the GLORIA observations. All major structures are reproduced, and weak indications are found again for the filaments and structures from 17:30 to 19:00 UTC. The overall ozone mixing ratios in the EMAC simulation are higher when compared to ICON-ART and closer to the absolute values observed by GLORIA.
The nitric acid distribution simulated by EMAC (Fig. 7h) matches the overall structure seen in the GLORIA data only qualitatively. Systematically lower mixing ratios are found in the EMAC data, and local maxima seen in the GLORIA observations between 14:00 and 19:00 UTC are barely reproduced. This is probably due to the fact that EMAC underestimates nitrification of the LMS in this particular winter. A similar underestimation of nitric acid simulated by EMAC was found for the Arctic winters 2009/2010 and 2010/2011, as discussed in Khosrawi et al. (2018), and also in the comparison to GLORIA measurements of research flight 21 on 18 March 2016, described in Khosrawi et al. (2017). However, the observed narrow filaments with low nitric acid reaching into the LMS between 17:30 and 19:00 UTC are again reproduced partly by the model.
Residuals between the model simulations and the GLORIA data are shown in
Fig. 7i–m. A systematic moist bias is seen in ICON-ART in the LMS, while
variable positive and negative residual patterns are found below the
tropopause (Fig. 7i). Note that there is hardly any variation in the
residual in the region of the narrow filaments above the tropopause from
17:30 to 18:30 UTC (Fig. 7i) due to the excellent agreement with GLORIA.
ICON-ART ozone (Fig. 7j) shows a systematic low bias above the tropopause,
while weak positive and negative residual patterns are found below the
tropopause. EMAC H
To investigate potential differences between the global R2B6 and the nested
R2B7 ICON-ART domain, differences between these grids are depicted in Fig. 8. Mesoscale patterns in the residuals of
Residuals between the ICON-ART nested R2B7 and global R2B6 domains
of
In summary, the dynamical situation is represented well by both models (with either consecutive ICON-ART forecasts or continuously nudged EMAC simulations) within the limitations of their horizontal resolution. Both models clearly reproduce the observed strongly subsided air masses in the western part of the flight and the narrow filaments between 17:30 and 19:00 UTC. Here, complementary patterns are found in the water vapour distribution when compared to ozone and nitric acid. Water vapour in the LMS is overestimated by EMAC, and ozone is underestimated by ICON-ART. Furthermore, EMAC clearly underestimates nitric acid and barely reproduces nitrification patterns seen in the GLORIA data.
Close-ups of the GLORIA, ICON-ART (nested R2B7 domain) and EMAC-STD trace
gas distributions are presented in Fig. 9. In Fig. 9a, two stronger moist
filaments reaching into the LMS up to
Close-ups of troposphere-to-stratosphere exchange region between 17:30 and 19:30 UTC. In the case of ICON-ART, the nested data are shown. For legend, see Fig. 7.
The ICON-ART simulation of specific humidity in the nested domain reproduces the vertical and horizontal extent as well as maximum mixing ratios very well (Fig. 9d). Even the weak filament in between the more developed filaments can be clearly identified. However, overall water vapour mixing ratios are slightly higher when compared to GLORIA. In the EMAC simulation, the two major filaments can be weakly identified, and warping of the dynamical tropopause is weaker (Fig. 9f). However, it has to be remembered that the horizontal resolution of the EMAC simulation is T106, which is lower than that of the ICON-ART R2B7 nest by about a factor of 5. Overall absolute water vapour mixing ratios are clearly overestimated by EMAC.
The GLORIA ozone distribution shows detailed fine structures close to the flight altitude. Structures low in ozone correspond to the high water vapour structures and extend further to flight altitude (Fig. 9b). The combination of ozone and water vapour data clearly shows that air masses characterised by tropospheric moisture levels reach deeply into the LMS and are connected to variations in the dynamical tropopause. Tropopause folds and steps in the tropopause are regions where isentropic levels cross the tropopause and jet streams. They are known bidirectional exchange regions between the tropopause and stratosphere (e.g. Shapiro, 1980; Keyser and Shapiro, 1986) and contribute to transport and mixing of tropospheric air into the LMS, as diagnosed by, for example, Werner et al. (2010), Krause et al. (2018) and Jing et al. (2018) (note however that a net exchange from the LMS to the troposphere dominates).
The simulation of ozone in the nested ICON-ART domain reproduces the same sequence of filaments, however with lower mixing ratios and less fine of a structure. EMAC reproduces the filaments around 17:30 to 18:30 UTC only faintly, while observed absolute mixing ratios are matched well. Finally, the GLORIA close-up in Fig. 9c shows a highly structured nitric acid distribution. EMAC again broadly captures the filaments, while mixing ratios are clearly underestimated and local maxima are barely reproduced (Fig. 9h).
In summary, Fig. 9 shows that ICON-ART using the R2B7 (
The evolution of the filaments seen in the GLORIA and model data is analysed
with the help of ICON-ART. Figure 10a, d, g and j show the horizontal
distribution of water vapour and horizontal wind from 23 until 26 February 2016 at 10 km altitude. The wind contours south of
Evolution of filaments in nested ICON-ART domain.
In the region of the moist upper-tropospheric air masses south of Greenland
and the evolving broad filament with low PV towards the pole on the
following days (Fig. 10k, h, e, b), the PV distribution shows meridional
overturning of the PV gradient that frames the moist upper-tropospheric air
masses. The pattern suggests poleward breaking of a cyclonically sheared
Rossby wave (e.g. Gabriel and Peters, 2008, and references therein). Thereby,
a separate isolated large patch of low PV values above western Greenland and
the Atlantic on 23 February 2016 (Fig. 10k) combines with the moist upper-tropospheric air masses with low PV in the south and seems to result from
another Rossby wave breaking event that had previously occurred. As a
consequence, a long broad filament with low PV stretches up to 80
The vertical cross sections shown in Fig. 10l, i, f and c correspond to the magenta lines in the left and middle column. The locations of the cross sections were chosen with the intention to cover the area sampled by GLORIA and to capture the connected atmospheric structures in the vicinity that are discussed above. As can be seen from the vertical cross sections shown in Fig. 10l, i, f and c, the evolving filaments are framed in the west and east by steep gradients in tropopause height. The larger moist filament originates from the region around the jet stream band that branched away during the Rossby wave breaking event (compare Fig. 10j, g, d, a). It is aligned nearly parallel to the 320 and 340 K isentropic levels on 23 February 2016 (Fig. 10l). At lower altitudes, the 300 K isentropic level crosses the dynamical tropopause in the west in Fig. 10l, i, f and c. As discussed by Shapiro (1980), such regions provide suitable conditions for bidirectional cross-tropopause exchange. At higher altitudes, the 4 PVU isoline crosses the 320 K isentropic level in the same region and suggests conditions suitable for isentropic transport across horizontal PV gradients also here.
Local oscillations of the isentropic levels on 23 February 2016 between 55
and 50
The other two filaments on 23 February 2016 in the east are associated with a tropopause fold remnant in the east (Fig. 10l). The tropopause fold remnant declines during the subsequent days, moves west (Fig. 10i, f) and joins with the newly formed tropopause fold in the west on 26 February 2016 (Fig. 10c). Since these two filaments are aligned steeply across the isentropic levels already on 23 February 2016, they are interpreted as older structures that were previously formed in a similar way like the stronger filament in the west.
Overall, the vertical cross sections in Fig. 10l, i, f and c show that the filaments observed by GLORIA evolved along steep gradients of the dynamical tropopause in connection with Rossby wave breaking. The larger filament in the west evolved during a Rossby wave breaking event, where moist tropospheric air masses were transported horizontally into the Arctic LMS along the jet stream under conditions suitable for cross-tropopause exchange. The other two filaments are interpreted as older structures in connection with a tropopause fold remnant in the east that probably evolved during a previous Rossby wave breaking event.
By scattering and correlating modelled mixing ratios with the observed
values, model discrepancies (and likely biases) can be quantified as
deviations from the ideal
Correlation of GLORIA H
For ICON-ART specific humidity, excellent agreement is found for high water vapour levels in the troposphere (Fig. 11a). At PV levels higher than
For ozone in the nested ICON-ART domain (Fig. 11b), a systematic low bias is
found that increases with PV. This is attributed to the simplified ozone
depletion parameterisation. For the T106 EMAC simulation the agreement in
ozone with GLORIA measurements is very good (Fig. 11d). Here, the data
points are well scattered around the
To quantify the simulated cumulative impact of ozone depletion and nitrification of the LMS in the ICON-ART and EMAC simulations during the entire winter until the flight date, corresponding passive tracers are simulated (Fig. 12). Residuals between the “active” tracers (i.e. chemical and microphysical processes activated) and the corresponding passive tracers (only dynamical processes act on them) indicate the cumulative net changes due to the processes considered in the “active” case.
Modelled ozone depletion and changes in nitric acid due
to chemical and microphysical processes. Residuals between
In the ICON-ART simulation, the “active” ozone tracer simulation shows
systematically lower mixing ratios than the “passive” ozone tracer (Fig. 12a) at all altitudes due to modelled ozone depletion. Above the dynamical
tropopause, the difference increases from
In the EMAC simulation (Fig. 12b), the residual is close to zero in the
troposphere, in the tropopause region and also at lower levels of the LMS.
Only in the deeply subsided vortex remnant around waypoint A and B is ozone
significantly lower in the “active” simulation, which is indicated by
residuals exceeding
While EMAC nitric acid agrees well with GLORIA in the troposphere, a
systematic low bias is found above the troposphere that strongly increases
with altitude (Fig. 11e). The bias amounts to
The EMAC nitric acid residual shown in Fig. 12c clearly shows that this
species is enhanced in the simulation by
Finally, the EMAC sensitivity simulations presented in Fig. 13 show that changing the model resolution from T106 to T42 exacerbates the LMS moist bias in the water vapour distribution (Fig. 13a; compare Stenke et al., 2008) and results in significantly lower mixing ratios in the LMS ozone (Fig. 13b) and nitric acid distributions (Fig. 13c) in the T42 simulation. A similar behaviour of EMAC was found in the stratosphere by Khosrawi et al. (2017), who stated that the T106 simulation agrees slightly better with Aura Microwave Limb Sounder observations for both species.
Modelled differences in H
Scavenging processes by cirrus cloud ice particles are capable of removing
trace gases from the gas phase. Sedimentation of the ice particles is
capable of removing the trapped gases from affected altitudes. While
previous studies focused mainly on scavenging on liquid cloud droplets (Tost
et al., 2010; Wang et al., 2010; Pierce et al., 2015; Kaiser et al., 2019),
Tost et al. (2010), however, found HNO
As can be seen in Fig. 13d–f, simulated scavenging processes result in
noticeable changes in the LMS only in the case of nitric acid. HNO
In the following, the diagnosed model biases and suggestions for model improvement are summarised:
We propose considering the model biases and deficits found here and our respective suggestions for future model development. As this work represents a case study, our findings hint at model deficiencies that might also be present in different seasons or latitudes. Further observations and model validation studies are needed to investigate these issues and to pinpoint these deficiencies to the respective deficits in the parameterisations.
Using GLORIA observations taken during the HALO long-range flight on 26 February 2016, we test the ability of the atmospheric chemistry model ICON-ART and the CCM EMAC to model mesoscale dynamical features, the chemical composition, and cirrus clouds and their impacts in the UT–LMS. The flight constitutes a multifaceted test case, covering deeply subsided air masses of the aged 2015/16 Arctic vortex, high-latitude LMS air masses, a highly textured region affected by troposphere-to-stratosphere exchange and high-altitude cirrus clouds.
In both models, even though very different in their character, the dynamical situation in particular, with the strongly subsided air masses in the western part of the flight, is simulated well. Here, the observed stratospheric air masses, characterised by low water vapour, high ozone and enhanced nitric acid mixing ratios, are reproduced.
The high-resolution ICON-ART set-up (in a consecutive short-forecast mode) involving an R2B7 nest (approx. 20 km) reproduces mesoscale dynamical structures also quite well. Narrow moist filaments in the LMS observed by GLORIA at tropopause gradients in the context of a Rossby wave breaking event and in the vicinity of an occluded Icelandic low are clearly reproduced by the model. A more detailed analysis with ICON-ART shows that a larger filament in the west was transported horizontally into the Arctic LMS in connection with a jet stream split during poleward breaking of a cyclonically sheared Rossby wave. Further weaker filaments are associated with an older tropopause fold in the east. Given the lower resolution of the nudged T106 simulation of the EMAC model, we find that this model also reproduces these features at the limit of the model resolution in a very reasonable way. All major cloud systems detected by GLORIA can be identified qualitatively in both models by cloud masks generated from the respective ice water content variables interpolated to the GLORIA geolocations. Remaining discrepancies between GLORIA and the models as well as between the two models are attributed to uncertainties in the modelled geolocations or timing of cloud scenarios as well as the limited qualitative comparison of the measured quantity cloud index with cloud masks generated from the models. We have demonstrated that residuals between the active water vapour tracer and the corresponding tracer neglecting cloud microphysics in the ICON-ART simulation can be used as an alternative proxy for the presence of clouds, in terms of an integrated picture of the short forecast. In particular, this proxy hinted at a cloud system observed by GLORIA from 14:00 to 15:00 UTC, which is not present in the ICON-ART simulation at this particular time. However, a corresponding cloud system is found in the model data a few hours prior to the measurement at this particular geolocation. Both models tend to simulate cloud systems reaching higher above the tropopause than observed by GLORIA and suggest that LMS humidity is significantly affected by cloud microphysics in the simulations. This is supported by the ICON-ART short-term sensitivity forecast neglecting cloud microphysics, which shows that LMS humidity can be depleted locally by cloud processes by 1–2 ppmv within less than 20 h.
Overall magnitudes of UT–LMS humidity are reproduced well by the consecutive ICON-ART short-term forecasts (reinitialised at 00:00 UTC with ECMWF IFS) and the continuous simulations of EMAC water vapour. However, a systematic moist bias is found in the LMS in both models. The same moist bias is known for the ECMWF and other weather and atmospheric forecast systems and is a contributing factor to a cold bias there in medium-range forecasts with these systems (Stenke et al., 2008). The fact that both models tend to simulate cirrus clouds reaching higher above the tropopause than observed by GLORIA might be related to the moist bias. Here, enhanced saturation versus the ice phase in the model simulations might be a reason for the cloud systems reaching to higher altitudes. Consistent with other studies (Roeckner et al., 2006; Polichtchouk et al., 2019), we find a higher moist bias in an EMAC simulation with a lower resolution (T42 instead of T106).
While the overall ozone mixing ratios of EMAC are in good agreement with
GLORIA, the simplified ICON-ART O
We find that LMS composition modelled by EMAC is notably affected by model
resolution. In addition to the enhanced moist bias, a reduction in
horizontal resolution from T106 to T42 leads to a low bias in ozone and an
even more pronounced low bias in nitric acid. This effect, concerning ozone
and nitric acid, has been also found in Khosrawi et al. (2017) when
compared to satellite data, however, with these authors focusing on higher
altitudes. These discrepancies might be overcome by resolution-dependent
model tuning. Finally, our EMAC simulations show that neglecting scavenging
processes by clouds has practically no impact on water vapour and ozone in
the LMS, while nitric acid is noticeably depleted by
Overall, we find that ICON-ART and EMAC T106 are well suited for comparison to high-resolution remote sensing aircraft data and are capable of simulating troposphere–stratosphere exchange in the context of Rossby wave breaking. Fine structures like the filaments seen in the GLORIA data between 17:30 and 18:30 UTC are reproduced well by ICON-ART and even modelled broadly by EMAC despite the much coarser resolution.
The GLORIA data were measured during a single flight on 26 February 2016
with a duration of 9 h 40 min and a total distance of
However, we find that accurate simulation of UT–LMS composition remains
challenging, and both models need to be further improved. We speculate that
the reported biases and sensitivities might help to provide better forecasts
and long-term projections by these and other models. The observed biases in
ICON-ART O
In this section we want to get back to the comparison of observed clouds by
GLORIA and modelled clouds by ICON-ART and EMAC. To prove that seemingly
“missing cloud systems” in the ICON-ART model, in particular the cloud
system from 14:00 to 15:00 UTC, had been present at some time prior to the
measurement at the respective geolocations in the model and to examine the
evolution of clouds during the day of PGS Flight 14, we have sampled the
model output of ICON-ART cloud variables (
Figures A1 to A3 show the evolution of clouds in the ICON-ART (panels d–f)
and EMAC (panels g–i) model at various times between
The cloud system detected by GLORIA from 14:00 to 15:00 UTC corresponds to geolocations along the westward flight leg between central Greenland and approximately the west coast of northern Greenland (see Fig. 1), with GLORIA pointing to the north.
Inspection of the panels d–f in Figs. A1 to A3 shows that a corresponding
cloud system is forming about 10 h before the measurement in the
ICON-ART model and is growing until it reaches its maximum vertical and
horizontal extent at about a time offset of
Afterwards (from
The corresponding cloud system in the EMAC simulation (Figs. A1–A3, panels
g–i) appears with a slightly different shape, but with remarkably larger
vertical extent, reaching down deep into the troposphere to about 6 km
altitude. A separate part appears close to flight altitude and seems to be
connected to the main cloud system in the troposphere. The connected cloud
system remains approximately constant from
In Sect. 4.1 we also find hints that the lower cloud system between 17:30
and 19:30 UTC, which was underestimated in the ICON-ART cross section, is
more pronounced at some time prior to the measurement. Inspection of Figs. A1 to A3 yields that the corresponding cloud system has been more developed
at these geolocations during the day of the flight, reaching its best
resemblance to the GLORIA cloud index around
However, we do not find any indications in Figs. A1 to A3 in the interpolated ICON-ART data (panels d–f) of a cloud located at waypoint A around 8 km altitude which would be responsible for the large cirrus cloud ice particle sedimentation signal in Fig. 6 and which is also visible in the EMAC data (cf. Fig. 5c).
In summary, this analysis yields that better resemblance of the ICON-ART cloud data to the GLORIA observations and EMAC simulations is found in some cases if model data of an earlier time step are considered.
In particular, the large cloud system observed by GLORIA from 14:00 to 15:00 UTC is reproduced in both the ICON-ART and EMAC model; however its vertical extent is much more pronounced in the EMAC model.
Both models show that this cloud system is subsiding with time, which is in accordance with the meteorological situation above central Greenland (a high-pressure system; cf. Sect. 3).
Same as Fig. 5, but model data (ICON-ART and EMAC) have been sampled with a constant time offset of
Same as Fig. 5, but model data (ICON-ART and EMAC) have been sampled with a constant time offset of
Same as Fig. 5, but model data (ICON-ART and EMAC) have been sampled with a constant time offset of
The data used here are available at the repository radar4KIT
(
FH and WW designed the study. FH and WW analysed the data with support from MH, PB, RR, SJ, AK, MW, OK, FK and JB. JB and MW performed the ICON-ART simulations. OK performed the EMAC simulations. HO, BMS and WW coordinated the HALO activities during PGS. FFV and the GLORIA team performed the GLORIA measurements and operations. JU, AK and SJ performed the GLORIA level-1 data analysis. SJ performed the GLORIA chemistry mode trace gas retrievals (level-2 data) used here. MH, AK, JU and SJ contributed to the discussion concerning the GLORIA cloud detection. FH and WW prepared the manuscript with contributions from the other authors. FH, WW, MW and FK designed the figures. All authors helped with discussions and with finalising the manuscript.
At least one of the (co-)authors is a member of the editorial board of
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “The Polar Stratosphere in a Changing Climate (POLSTRACC) (ACP/AMT inter-journal SI)”. It is not associated with a conference.
Atmospheric research with HALO is supported by the Priority Programme SPP 1294 of the Deutsche Forschungsgemeinschaft (DFG). The work of Florian Haenel has been funded by the DFG project no. 316735585 (WO 2160/1-1) and by the project “Advanced Earth System Modelling Capacity” (ESM) with project no. ZT-0003, funded by the Helmholtz Association. We thank the GLORIA team and DLR-FX for performing the measurements and HALO flights during PGS. The EMAC and ICON-ART simulations were performed on the supercomputer ForHLR and with the help of the Large Scale Data Facility at the Karlsruhe Institute of Technology, both funded by the Ministry of Science, Research and the Arts Baden-Württemberg and by the Federal Ministry of Education and Research. The interpolations of EMAC and ICON-ART simulation data to the GLORIA tangent geolocations were performed on the bwUniCluster (2.0). The authors acknowledge support by the state of Baden-Württemberg through bwHPC. We thank the ECMWF for providing the meteorological data used here. We further acknowledge support by the KIT-Publication Fund of the Karlsruhe Institute of Technology. We would like to thank the two referees and the editor (see Review statement) for their time and valuable comments.
This research has been supported by the Deutsche Forschungsgemeinschaft (DFG), DFG Priority Programme SPP 1294, within the DFG project no. 316735585 (WO 2160/1-1), and by the Helmholtz Association within the project “Advanced Earth System Modelling Capacity” (ESM) with project no. ZT-0003. The article processing charges for this open-access publication were covered by the Karlsruhe Institute of Technology (KIT).
This paper was edited by Mathias Palm and reviewed by Michelle Santee and one anonymous referee.