Introduction
The complexity of stratospheric air intrusions into the troposphere has been
investigated with lidar systems in great detail. A lot of information was
obtained from airborne transects (e.g. Browell et al., 1987, 1996, 2001;
Flentje et al., 2005) and ground-based time series (e.g. Ancellet et al.,
1991, 1994; Lamarque et al., 1996; Langford et al., 1996; Langford and Reid,
1998; Eisele et al., 1999; Stohl and Trickl, 1999; Galani et al., 2003; Zanis
et al., 2003; Trickl et al., 2003, 2010; Di Girolamo et al., 2009; Kuang et
al., 2012). Ozone is an excellent tracer for mapping intrusion layers, but
does not allow for the erosion of these layers within the troposphere to be
quantified because one cannot easily resolve the mixing of tropospheric air
into the descending layer. Water vapour is a much better choice for such
investigations, because of the low stratospheric volume mixing ratio of about
5 ppm (e.g. Scherer et al., 2008) and only slightly higher values just above
the tropopause.
Turbulent mixing has been identified as an important source of tropospheric
air in tropopause folds (Shapiro, 1976, 1978, 1980). About half of the air
mass in a fold has been estimated to be of tropospheric character (Shapiro,
1980; Vogel et al., 2011). Nevertheless, the tropospheric input had never
been quantified along the entire path of the air mass eventually reaching
the lower troposphere. An open question has been how much of the
tropospheric air originates already from the so-called “mixing layer” around
the thermal tropopause (e.g. Danielsen, 1968; Lelieveld et al., 1997;
Hintsa et al., 1998; Zahn et al., 1999, 2014; Fischer et al., 2000; Hoor et
al., 2002, 2004; Pan et al., 2004, 2007; Brioude et al., 2006, 2008; Sprung
and Zahn, 2010; Vogel et al., 2011) prior to the descent and how much of the
admixture occurs during the descent of an intrusion layer into the lower
troposphere. In some cases mixing of polluted or convectively lifted air
into intrusions within the free troposphere has been reported (e.g.
Parrish et al., 2000; Brioude al., 2007; Homeyer et al., 2011; Sullivan et
al., 2016).
In contrast to the idea of strong tropospheric mixing Bithell et al. (2000)
found in a case study that an extremely dry layer of presumable
stratospheric origin survived in the troposphere without resolvable change
for at least 10 days. Trickl et al. (2014, 2015) verified this behaviour
based on water-vapour measurements during about 80 intrusion cases: in 59 %
of the deep intrusion cases with subsidence times up to 6 days the
minimum relative humidity (RH) was 1 % or less, 1 order of magnitude
smaller than the typical results from in situ measurements with the
dew-point-mirror instrument at the nearby Zugspitze summit (2962 m a.s.l.).
The corresponding mixing ratio of roughly 50 ppm or less is typical of
values found in the “mixing layer” that extends a few kilometres into the
stratosphere.
Despite this evidence of low free-tropospheric mixing, the ozone number
densities in the same intrusion layers stay significantly below full
stratospheric values. Trickl et al. (2014) conclude that the ozone values
are mostly determined by how far the intrusion layer initially extends into
the stratosphere. They found that CO mixing ratios in deep intrusions rarely
strongly differ from tropospheric values. This implies that the descending
layers depart from the lowest few kilometres above the dynamical tropopause
since fully stratospheric CO values are substantially smaller.
Trickl et al. (2014) discussed three cases with rather filamentary structure
in order to demonstrate that exceptionally low mixing prevails even for thin
layers. In the follow-up paper presented here, sharing the main part of the
title, we extend that study by analysing a much more homogenous intrusion
layer over a rather large area: The observations took place over a major
part of central Europe during LUAMI (Lindenberg Upper-Air Methods
Intercomparison, in the evening of 17 October 2008; Wirth et al., 2009b).
Quantitative three-dimensional mapping with the DLR (Deutsches Zentrum
für Luft und Raumfahrt) airborne lidar system WALES (Water Vapour Lidar
Experiment in Space) (Wirth et al.,
2009a) around a major part of central Europe is combined with measurements
of ground-based lidar systems, balloon-borne sensors at four stations
forming the four corners of the flight track. Atmospheric transport
modelling shows the development of the descending dry layer between the
stratospheric source region over northern Canada and the Alps and clearly
confirms the ideas of the previous investigation. The campaign constitutes
one of the largest-scale comparisons of water-vapour profiling
instrumentation and verifies a very high quality of all the instruments
contributing. In particular, the first comparison of an airborne and a
ground-based differential-absorption (DIAL) system in the entire free
troposphere was made. Detailed results are given in the Appendix.
Results
Characterization of the stratospheric air intrusion on 17 October 2008 s
The intrusion was first detected in the routine forecast plot daily sent to
former STACCATO (Stohl et al., 2003) partner stations (Zanis et al., 2003).
Here, in Fig. 1 we give a revised version of that plot, now based on ECMWF
re-analysis meteorological data, and based on the all trajectories
calculated for the period between 8 October 2008, and 19:00 CET on 15 October 2008, fulfilling the criteria for deep stratosphere–troposphere
transport (STT) specified in Sect. 2.2. From these trajectories, Fig. 1 shows
those intersecting the 65∘ meridian between
60 and 75∘ N within ±6 h from 15 October, 01:00 CET. This time was chosen for a maximum horizontal extension
of the trajectory field towards Payerne that occurred during the second half
of 17 October when the measurements took place. The air pressure range of
the main layer over central Europe is roughly 650 to 750 mbar (about
2.5 to 3.7 km), the lower altitudes being found to the south-west, the
higher ones to the north-west of the trajectory bundle, in agreement with
the well-known fold structure transverse to the flow (e.g. Danielsen,
1968).
(a) Slant view from the north-east of the WALES flight in
the late afternoon of 17 October 2008: P means Payerne (Switzerland), B
Bilthoven (the Netherlands), L Lindenberg (north-eastern Germany), and Z
Zugspitze (Garmisch-Partenkirchen, southern Germany). The flight track (red
line) is shown above the “curtain” of the water-vapour profiles from the
airborne DIAL measurements. The track of the Lindenberg sonde (S) is marked
by a black line. (b) Vertical distributions of water vapour (bottom)
and 1064 nm aerosol backscatter ratio (top) along the flight track; the times
of the overflights of the four stations are marked by red vertical lines.
![]()
The intrusion arrived over central Europe following a frontal system with
rain that passed over the eastern Alps to the south-east during the
preceding night (not shown). A period of clear weather started, which was associated
with the arrival of a high-pressure zone. The water-vapour images of the
geostationary satellite METEOSAT show just moderate drying after the frontal
passage. A slightly drier, hook-shaped feature arrived over northern Germany in
the morning of October 16. It moved eastward to Poland until the following
day. No indication of the intrusion is seen further to the south. However,
these images are more representative for the upper troposphere. As will be
shown below, the dry intrusion layer proceeded well hidden in a rather moist
middle and lower troposphere.
DLR measurements
The flight path of the DLR Falcon jet is marked in Fig. 1. Colour-coded
summary plots of the measurements during the flight on 17 October 2008 are
given in the panels of Fig. 2. The flight started at Oberpfaffenhofen (ICAO
(International Civil Aviation Organization) code EDMO) at 16:42 CET. The
aircraft turned to the west and climbed to about 11 km altitude. It first
reached Payerne at 17:18 CET, then Bilthoven at 18:15 CET, Lindenberg at
19:03 CET, and finally Garmisch-Partenkirchen (Zugspitze) at 19:52 CET The
data are most accurate in the upper troposphere, i.e. close to the
aircraft, but are remarkably reliable even in the lower troposphere, where
the lidar signal is much weaker and, thus, noisier. The relative noise level
within the dry layer additionally grew whenever the water-vapour density
above the intrusion was enhanced to an extent that much of the radiation was
absorbed. The lower-tropospheric performance was, thus, the best over
Lindenberg (see Fig. 2) and becomes evident from the comparisons that are
shown in the Appendix, with one exception.
In the lower panel of Fig. 2 also the backscatter ratio for 1064 nm is
given, i.e. the ratio of the total backscatter coefficient and the Rayleigh
backscatter coefficient. Any value exceeding 1.0 means the presence of
aerosol, and high values around the upper end of the scale can be attributed
to clouds. The data gaps (white areas) are mostly associated with the
presence of clouds at the top of the boundary layer or cirrus clouds and the
corresponding light loss.
Quite importantly, slightly enhanced aerosol was retrieved in the upper half
of the intrusion layer along the entire flight path. The most reasonable
explanation of this observation would be a downward transport of some of the
enhanced stratospheric aerosol after the violent eruptions of Okmok and
Kasatochi (to 15 and 13.7 km, respectively; Massie, 2015) starting on
12 July 2008 and on 7 August 2008, respectively, which was also registered
with the stratospheric aerosol lidar at Garmisch-Partenkirchen (Trickl et
al., 2013), up to about 19 km in October 2008. More information on this
remarkable observation can be found in some of the following
sections.
Payerne
During the hours of the LUAMI campaign Payerne was located close to the
western edge of the intrusion layer (Fig. 1). Nevertheless, the time series
of the Raman lidar (Fig. 3a) verifies the presence of a very dry
layer between 2 and 3 km during the entire period displayed, starting at
13:45 CET. The driest period with mixing ratios of 35 to 65 ppm started at
about 17:30 CET (Fig. 3b); 50 ppm is a typical value as found in
the tropopause region (Trickl et al., 2014). The relative uncertainties of
the minimum mixing ratios specified for the period before 17:00 CET are 7 to
19 %, after 17:00 CET 5 to 9 %.
The presence of stratospheric air is confirmed by the 13:00 CET ozone
profile (Fig. 4) that exhibits a 76.2 ppb maximum at 3.2 km, residing on a
background of roughly 50 ppb. It is interesting to note that the
corresponding RH minimum is downward shifted by about 0.3 km. The midnight
(00:00 UTC or 01:00 CET) RH minimum was 1 %, presumably a truncation value
(Trickl et al., 2014). This low value is in agreement with the drier
situation revealed by RALMO for the night.
Bilthoven
The time series of the CAELI system is depicted in Fig. 5. The noise at
early times is due to clouds passing over the lidar. Two dry layers are
visible. However, the minimum mixing ratios are of the order of 500 ppm
(Fig. 6), which is beyond typical values in the lowermost stratosphere. By
contrast, the noon sonde measurement at De Bilt (KNMI) (Fig. 6), as in the
case of Payerne, shows the typical low-humidity cut-off at 1 % RH (about
70 ppm). Even 70 ppm are, again, within the range of values frequently found
just above the tropopause. It seems that at the time of the lidar
measurements in Fig. 5, the driest part of the intrusion was already over.
Around midnight, the intrusion layer had almost disappeared as can be
concluded from the 24:30 CET sonde measurement.
The two lidar systems agree well in a range up to 8 km (Fig. 6). There are
just a few exceptions outside the specified uncertainties most likely due to
insufficient spatial matching, or far-field detection of the DLR lidar. The
agreement with the sonde data is not satisfactory due to the considerable
time differences, except for the range between 3.2 and 7 km in the midnight
profile.
In addition, a profile from the ECMWF analysis is shown. Outside the dry
layers the agreement is reasonable, but just one of the two layers seen in
the measurements is indicated. Another example can be found in the Appendix
(Payerne, Fig. A1).
(a) Water-vapour time series of RALMO on 17 October 2008;
the time of the aircraft overflight (17:18 CET) is marked by a red vertical
line. The figure is based on 0.5 h averages, the times being centred in the
respective measurement interval. Before 18:15 CET the residual daylight
background prevented measurements beyond the lower free troposphere.
(b) Time series of the water-vapour minimum in the stratospheric
intrusion layer on 17 October 2008, as recorded by RALMO, and the
corresponding altitude; the time of the aircraft overflight (17:18 CET) is
marked by a red vertical line. The figure is based on 0.5 h averages, the
times being centred in the respective measurement interval.
![]()
Lindenberg
During the campaign the lidar data of RAMSES were prepared as 10 min and
30 min averages. The time series of the 30 min data are shown in the upper
panel of Fig. 7, also using some 10 min data next to the data gaps. The
measurements were continued until 06:00 CET on 18 October. During the period
displayed the intrusion layer became continually thinner. The data do not
exhibit a single minimum of the mixing. In the lower panel of Fig. 7, we,
therefore, show the minimum values for the two driest zones in the upper
panel separately. The minimum mixing ratios retrieved are 120 ppm, which is,
still, in some agreement with conditions inside the “mixing layer” of the
tropopause region (Trickl et al., 2014), but clearly higher than the minima
observed at the other sites. The relative uncertainties of the RAMSES mixing
ratios specified in the vertical range around the intrusion are just a few
per cent.
The ozone profile measured by the balloon payload launched at 18:44 CET is
shown in Fig. 8. Quite interestingly, the highest ozone peak (75 ppb) was
observed at the upper end of the dry layer at an altitude of about 5.5 km,
although just 0.1 km above the RH minimum (5 %). This is in agreement
with the idea that the ozone rise in the lowermost stratosphere of the
Arctic source region was transferred to Lindenberg without major change,
assuming low interference by tropospheric air during the transport (Trickl
et al., 2014). A similar behaviour is indicated for Payerne in Fig. 4.
The ozone structure above 6 km is not clear. There is an obvious
anti-correlation of ozone and RH indicating stratospheric influence.
However, the elevated RH values could indicate mixing with tropospheric air.
Sonde ozone and relative-humidity profiles above Payerne on
17 October 2008; the times are launch times.
![]()
Water-vapour time series of CAELI on 17 October 2008 (Bilthoven);
the time of the aircraft overflight (18:16 CET) is marked by a red vertical
line. In the graph, only the data from the far-field receiver are shown
(above 1.7 km). The profiles are shown at the full native resolution of
10 s and 7.5 m.
![]()
In Fig. 9a the colour-coded plot of the water-vapour mixing ratio, derived from
the radiosonde ascents at Lindenberg between 14 and 23 October, is shown. The
plot benefits from the 6 h intervals between the launches at
Lindenberg, shorter than the conventional 12 h. If one neglects
uncertainties due to the graphical procedure applied, there is a strong hint
on a direct connection of the dry layer to the stratosphere during the first
half of 17 October (Julian day 291) that is also indicated in the upper
panel of Fig. 7. The transverse drift of the fold away from Lindenberg is
confirmed by the model calculations (Sect. 3.7).
The aerosol backscatter coefficients derived from the 354.84 nm RAMSES
measurements are rather noisy due to the very strong contribution from
Rayleigh backscattering at this short wavelength. Nevertheless, a small
spike (backscatter ratio 1.05) is seen in the profile next to the DLR
overflight at 5.08 km (not shown), residing on a broader pedestal between
3.8 and 5.2 km. This structure is in good agreement with the WALES results
(Fig. 2). The result of a 3-h average is shown further below (Sect. 3.6).
Comparison of the water-vapour mixing ratio from the CAELI Raman
lidar (18:23 CET, i.e. 15 min average between 18:15 and 18:30) and the
airborne lidar (18:15 CET) at Bilthoven; the mixing ratios from the routine
noon and midnight measurement at De Bilt (station code 6260) are given for
comparison. The noon profile reveals a much more pronounced stage of the
intrusion than the lidar data. In addition, the humidity result from a
high-resolution ECMWF analysis for the time of the aircraft arrival is shown,
again just indicating the intrusion layers (one of the two). The times for
the lidar systems refer to the middle of a measurement, for the sonde the
launch time (LT) was taken.
![]()
(a) Water-vapour time series of RAMSES on 17 October 2008
(Lindenberg); the time of the aircraft overflight (19:03 CET) is marked by a
red vertical line. (b) Time series of the two water-vapour minima in
the stratospheric intrusion layer on 17 October 2008, as recorded by RAMSES,
and the corresponding altitude; the time of the aircraft overflight
(19:03 CET) is marked by a red vertical line. The figure is based on 0.5 h
averages, the times being centred in the respective measurement interval.
![]()
Zugspitze
On 17 October 2008 a total of five measurements with the water-vapour DIAL
at UFS were made between 16:55 and 20:55 CET. Figure 10 gives an overview of
the profiles. The data are given as number densities, which is the primary
quantity measured by DIAL systems (not requiring the additional use of sonde
data). During that time period the intrusion layer descended by about 0.6 km. The minimum densities ranged between -7.7×1019 m-3
(the negative value being caused by data noise) and 7.9×1020 m-3 m-3, with a standard
deviation of 7×1020 m-3 (corresponding to a mixing ratio of roughly 37 ppm). The figure
suggests that the intrusion cut a descending dry hole into a triangular
humid distribution that was gradually restored, as indicated by the growing
peak density.
The noon and midnight RH profiles of the Munich (Oberschleißheim, WMO
station 10868, 100 km roughly to the north) sonde type RS92 extend the
range of descent over southern Bavaria to 3.9 km (thick red arrow in Fig. 10) → 2.78 km. Because of the complexity of Fig. 10 we do not include
the corresponding H2O density profiles there. We used high-resolution
data received from the German Weather Service (DWD). In this data set, all
four minima for Munich and Stuttgart between noon and midnight reach the
cut-off value of 1 % RH. The situation seems to differ from that in
Lindenberg: the tropopause for the preceding ascent (01:00 CET on 17 October)
does not exhibit a strong lowering and the RH values are rather high
throughout the troposphere. However, the time difference of 12 h is too long
to be absolutely sure about excluding a direct connection of the dry layer
to the stratosphere over southern Germany.
It is interesting to note that, despite uncertainties of the sonde results,
the value of 1 % RH has been found to be quite typical in the routine
analyses of STT events at Garmisch-Partenkirchen since 2007. This value is
clearly dominating for low to moderate travel times. For subsidence times
beyond 10 days, the RH minima may grow to 2–6 %. In the current study 1 % RH was consistently observed in the sonde data in the vicinity of all
sites involved but Lindenberg, where also the lidar minima are slightly
higher.
Ozone and RH profiles during the sonde ascent launched at Lindenberg
at 18:44 CET; for the RH of the RS92 sonde the final GRUAN data product was
taken.
![]()
Time series of the water-vapour mixing ratio over Lindenberg during
the period between 14 (Julian day 288) and 23 (Julian day 297) October 2008;
this figure was derived from radiosonde ascents at intervals of 6 h. The
mean flight times (in UTC) are marked by arrows. The thermal tropopause is
indicated by a black line. The intrusion examined in this study is visible on
Julian days 291 (17 October) and 292. The graphics indicate the passage of a
major part of the tropopause fold over Lindenberg.
![]()
The Zugspitze in situ measurements showed a drop in relative humidity right
after the end of the lidar comparison in agreement with further descent of
the dry layer (Fig. 11). The minimum half-hour average, 7.2 %, was not
reached before 01:00 CET, which indicates considerable slowing of the
subsidence. A pronounced ozone rise to more than 73.3 ppb was found that
started 4 h later than the beginning of the humidity drop. Both the
peak ozone value and the delay are in agreement with the findings for Payerne
and Lindenberg where, within the intrusion layer, an ozone rise towards
higher altitudes was found. This observation must be considered in future
data-filtering efforts of the half-hour averages for quantifying the
stratospheric fraction of the Zugspitze ozone as described by Trickl et
al. (2010). Carbon monoxide stayed above 110 ppb, which is rather typical
and, again, indicates that the air mass originates just in the tropopause
region (Trickl et al., 2014). Fully stratospheric CO values are substantially
lower.
Water-vapour profiles from the measurements of Zugspitze DIAL on
17 October 2008; the red arrow marks the vertical position of the RH minimum
(1 %) from the noon “Munich” radiosonde, observed during the ascent at
11:56 CET. The grey dashed line marks a mixing ratio of 100 ppm as
determined from the same sonde ascent. In addition, a density profile for
75 % RH is given for a crude comparison (Munich, 13:00 CET). The
corresponding profile for the following midnight shows significantly higher
RH below 4.5 km since the intrusion had subsided to 2.78 km and, therefore,
is not included here.
![]()
Zugspitze in situ measurements of ozone, carbon monoxide and
relative humidity on 17–19 October 2008; the stratospheric layer is clearly
visible in the H2O and O3 data, but there is no significant hint in
the CO curve.
![]()
As in the 1064 nm measurements of WALES aerosol the UFS measurements show
aerosol in the upper half of the intrusion layer, with a peak 817.2 nm
backscatter coefficient of about 1.47×10-7 m-1 sr-1
(Fig. 12; the corresponding backscatter ratio is 1.76). The aerosol
structure could be so clearly detected by both DIAL systems because of the
low noise of the rather small Rayleigh background at the long wavelengths
used. We show two examples, one profile in the late afternoon (16:55 CET)
that contains the entire aerosol peak centred at 3.57 km, but ending below a
cirrus layer, and another profile around the time of the comparison when the
aerosol peak was located at the lower edge of the useful range of the
backscatter profile. The second profile was also evaluated in the
stratosphere and shows two peaks of the volcanic eruptions (at about 12.7
and 16.2 km), in addition to the stratospheric background that extended from
the tropopause to about 25 km during the background phase preceding the
eruptions (Trickl et al., 2013). The stratospheric peaks are considerably
smaller than the peak inside the intrusion layer because of the much lower
atmospheric density. This explanation assumes that a comparable aerosol
density was also present over the source region, which looks reasonable many
months after the eruptions. However, full homogeneity was not reached as
seen in the figure and as was discernible in the aerosol profiles of the
DIAL that varied from hour to hour on that day. In the lower panel of Fig. 12 an expanded section of the profile for 17:02 CET is given, together with
the corresponding water-vapour profile rescaled to fit into the plot window.
As already concluded from Fig. 2, the aerosol peaks in the upper half of the
intrusion layer, where also the lowest humidity is found. This resembles the
behaviour of the ozone distribution at Payerne and Lindenberg.
(a) 817.2 nm aerosol backscatter coefficients for two of
the measurements at UFS on 17 October 2008; the two spikes between 3 and
4 km are located inside the intrusion layer at these times. For comparison,
a 354.84 nm profile from a 3 h average of RAMSES measurements around the
time of the DLR overflight of Lindenberg is shown, rescaled for 817.2 nm.
Here, the corresponding aerosol peak was detected at about 5 km. In the
stratosphere two volcanic aerosol layers related to the Okmok and Kasatochi
eruptions are seen. (b) Vertically zoomed 16:55 CET measurements at
UFS: 817.2 nm backscatter coefficients and water-vapour density, scaled to
fit horizontally into the frame.
![]()
In addition, the 3 h average for Lindenberg around the overflight time is
inserted into the upper panel in grey colour. The curve is rescaled by
multiplying the values with 354.84/817.2 according to a λ-1.4
Ångström law guessed from the curves of Jäger and Deshler (2002)
for the wavelength dependence of the backscatter coefficients. The
Lindenberg aerosol peak inside the intrusion is located at 4.97 km
(2.3×10-7 m-1 sr-1, backscatter ratio 1.05), i.e.
slightly downward shifted due to the long averaging. The tropopause above
Lindenberg was at just 10.1 km, which explains the lower position of the
lower volcanic layer just above this altitude.
Right panels: intrusion trajectories (black lines) intersecting a
vertical control surface (green line) above northern Canada at three
different times; left panels: intersection points of the trajectories on the
control surface (see text); the blue contour lines are isentropes (in K).
![]()
Transport modelling
The 5-day trajectories were released and preselected for deep subsidence from
the lowermost stratosphere as described in Sect. 2.2. In the next step, cross
sections transverse to the flow were prepared at a number of locations
between Canada and the Alps. Examples for four of the locations are shown
here. The PV contours (in colour), isentropes (as blue contour lines), the
interpolation points of the individual trajectories closest in time (within
±6 h) to the cross sections (yellow dots), and the points of
intersection of the trajectories (magenta dots) are displayed. The cross
sections allow the
position of the air parcel to be seen relative to the dynamic tropopause
(2 pvu isosurface) and highlight their way down from the stratosphere to the
lower troposphere. Note that all trajectories calculated, i.e. starting
between 8 and 15 October and fulfilling the deep-intrusion criteria in
Sect. 2.2, contribute to the cross sections.
Right panel: intrusion trajectories (black lines) intersecting a
vertical control surface (green line) above the west coast of Greenland at
three different times; left panel: intersection points of the trajectories on
the control surface (see text); the blue contour lines are isentropes (in
K).
![]()
The temporal development of the tropopause and positions of the trajectories
as they cross a first vertical surface (along 80∘ W, green
line) north of Hudson's Bay are shown in Fig. 13. The figure
shows three examples selected from the full time period 14 October, 06:00 UTC,
to 15 October, 12:00 UTC, representing the phase of the highest trajectory
density at this cross section and the first indication of diminishing. The
trajectories (black lines) are shown in the right panels.
At the location of the transverse surface of Fig. 13, the tropopause is only
slightly distorted toward lower altitudes, during the entire period covered.
The beginning of the trajectories selected by the deep-STT criterion stays
east of 100∘ W, i.e. well inside the model domain. This
means that the chosen control surface is close to the true beginning of the
intrusion. The tropopause is located clearly below the minimum pressure
level of 200 mbar. Please, note that the trajectories concentrate not far
from the 2 pvu surface. This nicely confirms the conclusion of Trickl et al. (2014)
from the rather high Zugspitze CO values in intrusions (see also Fig. 11) that the intrusions emerge from a shallow layer just above the dynamical
tropopause.
In the next cross section farther downstream (50∘ W, Fig. 14), already a fully developed tropopause fold is seen. The highest
trajectory density in the vicinity of this cross section was found between
06:00 and 18:00 UTC on 15 October. Here, we select the situation for 12:00 UTC
as an example for which also the lowest position (about 520 mbar) of the
2 pvu contour and the most pronounced westward extension of the intrusion
over central Europe (corresponding to the most pronounced dryness over
Payerne) were obtained. It is interesting to note that for the entire 12 h
period of maximum stratospheric density, the trajectories intersect the
control surface above the centre of the fold. The lowest deviation from the
centre was found for 06:00 UTC.
The best coincidence with the next transverse surface at 30∘ W (western Iceland, not shown) was calculated for the period 18:00 to
24:00 UTC on 15 October, the lowest position of the dynamical tropopause (about
500 mbar) occurring at 06:00 UTC on 16 October. However, at this time the
trajectory dots were positioned even fully above the fold.
The next surface was selected from 50∘ N, 0∘ E to 54∘ N, 20∘ E, approximately
representing Bilthoven and Lindenberg (shifted less than 1∘ to the north). In Fig. 15 we
show the panels for 12:00 to 24:00 UTC on 17 October. The trajectories cover Bilthoven in the first two right panels, but
move eastward towards midnight. This is in qualitative agreement (though
slightly later) with the rising minimum humidity in the observations. Over
Lindenberg, the trajectories seem to confirm the extended vertical range
(roughly 800 to 600 mbar) seen in the lidar measurements. In addition, the
sequence of panels shows an eastward propagation of the fold along the
control surface, in agreement with the radiosonde measurements shown in Fig. 9.
The trajectories in Fig. 15 pass east of Payerne. Those covering Payerne
reach the coastal area 6–18 h earlier (not shown).
Right panels: intrusion trajectories (black lines) intersecting a
vertical control surface (green line) along the coast of the Netherlands,
Germany, and Poland at three different times; left panels: intersection
points of the trajectories on the control surface (see text); the blue
contour lines are isentropes (in K).
![]()
Finally, a cross section slightly north of the Alps (44.5∘ N, 2∘ E to 51∘ N, 18∘ E) was
prepared (Fig. 16), almost exactly hitting Payerne and passing
0.9∘north of Garmisch-Partenkirchen/Zugspitze. We cut off
the cross section to the north-east, not reaching the end of the intrusion.
This decision was made because a number of stratospheric trajectories from
outside the trajectory field in Fig. 1 (not intersecting the first control
surface at 80∘ W) are located there and adds complexity.
Thus, we cannot judge if the layer starts to detach from the fold. In any
case, the radiosonde measurements at Payerne and Munich do not indicate a
connection of the intrusion layer to the stratosphere as in the case of Fig. 9. However, the radiosonde ascents at these stations took place at longer
intervals of 12 h, perhaps too coarse to see more details.
In Fig. 16 we give three examples of model calculations again for 12:00 to
24:00 UTC on 17 October. During this time the best overlap of the trajectories
with Payerne is found, in agreement with the growing dryness observed during
this period. Due to the cut-off towards the north-east (mentioned above) the
trajectory dots do not reach the high-PV contours, which is the case for a
longer control surface.
Right panels: intrusion trajectories (black lines) intersecting a
vertical control surface (green line) north of the Alps at three different
times; left panels: intersection points of the trajectories on the control
surface (see text); the blue contour lines are isentropes (in K).
![]()
It is obvious that the trajectory dots in the cross sections downstream the
intrusion exhibit a higher spread. To some extent this is ascribed to the
higher temporal jitter and to additional stratospheric contributions from
outside the main descending air stream. There is not a perfect matching of
the dots with the vertical contour of the fold for the earlier times. Later,
during the driest phase observed over Payerne (lower two panels), there is a
better agreement of the central axes. However, the trajectories for the
beginning of 18 October no longer horizontally overlap with the Swiss
station as can been judged by comparing the green bars in the right panels
of Fig. 16.
Discussion and conclusions
There is growing evidence that ozone injection from the stratosphere is very
likely a much stronger source of tropospheric ozone than frequently thought
(e.g. Roelofs and Lelieveld, 1997; Trickl et al., 2010, 2011, 2014).
However, a quantification of STT remains a difficult task. The results
presented in this paper, together with the findings of the preceding studies
(Trickl et al., 2014, 2015), are an important prerequisite on the way to
quantifying STT based on observational data alone, at least at a few
suitable stations: the low concentrations of water vapour found in most deep
stratospheric intrusions examined suggest that the intrusion layers reach
high-lying atmospheric observatories with rather little modification during
the transport. Thus, the long-term observations of ozone, RH and 7Be at
these stations can, therefore, yield a reasonable estimate of the impact of
STT at these sites (Stohl et al., 2000). Based on data filtering Scheel (2005) estimated of the annual fraction of STT ozone and the 1978–2010 trend
in the Zugspitze ozone related to STT. A revision of this effort is now
planned, based on the methods derived by Trickl et al. (2010). Still, an
approach of treating the initial tropospheric component of the air mass in
the tropopause region must be developed (Trickl et al., 2014).
As pointed out in the earlier paper (Trickl et al., 2014) the detection of
low free-tropospheric mixing means a considerable challenge for atmospheric
modelling, particularly for narrow layers. As demonstrated by Roelofs et
al. (2003), a very high spatial resolution is required for obtaining
reasonable trace-gas distributions. Limitations in Eulerian models are
imposed by numerical diffusion (Rastigejev et al., 2010).
The LUAMI measurements on 17 October 2008 have made a thorough
comparison of different high-quality instruments for water-vapour sounding possible,
in particular the CFH sonde, differential-absorption, and Raman lidar
systems. The airborne lidar served as a transfer standard. With respect to
the intercomparison of the instruments, the following main conclusions can
be drawn:
Apart from a generally excellent mutual agreement of the systems a high
capability of determining very low humidity levels was verified such as
those needed in the current study. The RS92 radiosonde (e.g. Miloshevich et
al., 2006; Vömel et al., 2007b; Steinbrecht et al., 2008; Dirksen et
al., 2014) was verified to reproduce RH values around 1 % indicating a
capability of resolving even lower values. The ground-based lidar systems
were found to resolve significantly lower humidity in the deep stratospheric
air intrusions since these layers are measured at relatively short
distances.
The campaign was to a major extent based on lidar measurements. Lidars are
ideal due to the important (Vogelmann et al., 2011, 2015) advantage of volume
matching and of producing dense time series. Since the airborne DIAL provided
information of the spatial structure of water vapour also the quality of the
balloon-borne instruments could be judged. At Lindenberg the spatial matching
of WALES and the balloon was particularly good since the aircraft flew along
the wind direction.
The signal of Raman lidar systems (Payerne, Bilthoven, and Lindenberg) is
proportional to the H2O density divided by the square of distance. As a
consequence, these systems are advantageous for measurements under very dry
conditions, at least during night-time. Without a noisy solar background the
humidity determined from Raman lidar systems is positive from its very
principle since each signal photon is caused by backscattering of the laser
radiation by H2O. At least after 18:00 CET on 17 October the
measurements of the Raman lidar systems are invaluable for this study since
they yield very reliable values for the humidity minima in the intrusion
layer above the respective site (uncertainty: roughly 5 ppm H2O).
The water–vapour data of DIAL systems in dry layers are noisy at all times
since they are based on absorption measurements in a noisy backscatter
signal. Under very dry conditions the noise can lead to pointwise negative
humidity values. However, as concluded previously (Trickl et al., 2014), the
comparisons confirmed that also DIAL systems can rather reliably determine
low values: As found by Trickl et al. (2014), the minimum uncertainty of the
ground-based Zugspitze DIAL in dry layers in the lower free troposphere
under optimum conditions is of the order of 25 ppm (roughly 5×1020 m-3 as to density, or 0.5 % RH).
With respect to the dynamics of the intrusion some interesting findings
could be found based on the measurements and the modelling study. In
particular, the observations, carried out in a rather wide region, together
with the model calculations have led to a thorough characterization of the
intrusion. The cross sections prepared with LAGRANTO trajectories nicely
show the development of the main intrusion layer from the source region in
arctic Canada on its way to the Alps. The main conclusions are
The minimum water-vapour mixing ratios observed above most sites
participating were clearly below 100 ppm during the driest periods, the
lowest values having been about 35 ppm (Payerne) or less (Zugspitze). The
dryness above Payerne is a remarkable fact since this station was close to
westernmost edge of the intrusion. The low values harden the conclusions of
Trickl et al. (2014) that significant mixing of the stratospheric air during
the downward transport to 3–4 km takes only place if there is external
interference from nearby frontal systems or convection. Stratospheric air
layers can travel over very long distances without losing much of their
characteristics (Trickl et al., 2014, 2015). Sometimes they survive with
minor mixing even when travelling once around the globe (Trickl et al.,
2011).
The formation of the tropopause fold has started significantly earlier and at
slightly higher altitudes than anticipated from the daily forecasts received
since autumn 2000 (Zanis et al., 2003). However, the success of the forecasts
(Trickl et al., 2010) could be due to the fact that a minimum start pressure
of 250 mbar (about 10.5 km) stays within the range of lowered tropopause
positions in the start region of the folding, even in summer.
The trajectory bundles transversely propagating in the folds initially stay
rather narrow, narrower than the fold structure marked by the PV = 2 pvu
contours. Later on, the bundle seems to expand, to some extent due to the
temporal spread at the control surfaces and due to additional STT
contributions entering the cross sections. Acute-angled fold structures have
repeatedly been observed with the ozone DIAL over Garmisch-Partenkirchen
(e.g. Trickl et al., 2010).
Downward motion occurs all along the path, initially faster on the west
side.
The changes in trajectory density and position shifts of the trajectory
bundle qualitatively confirm the time periods of the driest parts of the
layer in the observations at the different sites.
The LUAMI measurements of water vapour, ozone, and aerosol have indicated
another behaviour of descending stratospheric layers. As hypothesized by
Trickl et al. (2014) the ozone and aerosol distributions in the intrusion
layer is in agreement with the idea of a rather unperturbed transfer of the
vertical distribution of these species in the source region to Europe: An
increase of ozone from tropospheric values at the bottom of the layer to
elevated values at near the top of the intrusion was documented at three
stations, the location of the aerosol peak in the upper part for the entire
DLR flight. The straight air flow out of the lowermost stratosphere revealed
by the model calculations (transverse to the fold) confirms this idea. More
cases must be analysed to harden these findings.
As in the vast majority of the ozone observations with the lidar at
Garmisch-Partenkirchen the peak O3 mixing ratio in the parts of the
intrusions reaching the lower troposphere are moderate (60–80 ppb). Most
intrusions originate in the lowest layer above the dynamical tropopause (or
just slightly below) as concluded by Trickl et al. (2014) from the very small
drop in Zugspitze CO that is reproduced in Fig. 11. This is now clearly
verified by the modelling results (Fig. 13). In the middle troposphere the
layers are wider and the ozone mixing ratio ranges between 80 and 150 ppb
(e.g. Stohl and Trickl, 1999; Trickl et al., 2010). This indicates that
farther to the north in the source region (Fig. 13) the exiting layer extends
deeper into the stratosphere. Exceptions are rare. For example, on 1 October
2015 a layer 6 km wide with up to 235 ppb of ozone was registered with the
ozone lidar, between 5 and 11 km a.s.l. However,
this is, still, far away from peak ozone mixing ratios of the order of 5 ppm
found in the stratosphere above 20 km.
The aerosol seen in Figs. 2 and 12 in the upper half of the dry layer seems
to reflect the behaviour of ozone, which increased backscatter coefficients
towards the layer top. It is reasonable to assume that the lower volcanic
peak was located just above the tropopause in a major part of the Northern
Hemisphere. Stratospheric aerosol in intrusion layers has been rarely
reported (e.g. Browell et al., 1987; Langford and Reid, 1998). We have seen
indications in the ozone plus aerosol soundings at Garmisch-Partenkirchen in
2009 following the Sarychev eruption, or after a 1991 pyro-cumulonimbus in
the Québec province of Canada (Carnuth et al., 2002; Fromm et al., 2010).
STT has been seen as the most important removal mechanism in the
mid-latitudes, limiting the stratospheric dwell time of aerosol in the
mid-latitude stratosphere to 1 year and less (Trickl et al., 2013). As a
consequence, also the stratospheric impact of boreal smoke plumes (e.g. Fromm
et al., 2008, and Fig. 1 of Trickl et al., 2013) or particle formation from
aircraft emissions at high cruising altitudes strongly diminishes within less
than half a year.
In Figs. 6 and A1 (see below) high-resolution ECMWF profiles are presented.
These profiles were calculated for the entire flight track (Wirth et al.,
2009b). The ECMWF analysis shows roughly the same H2O distribution as
the measurements, but (as in Figs. 6 and A1) it is apparent from its much
smoother structure that the model is by no means able to resolve the fine
structure of the dry layer. The mean deviation between the WALES measurements
and the ECMWF analyses is -13 % (i.e. WALES is dryer). If the altitude
region of the dry layer is excluded, the mean difference is about -8 %.