Introduction
In mountainous terrain the atmospheric boundary layer, and thus
the living conditions in these regions, are governed by processes of
different scales. Under fair weather conditions, the atmospheric boundary
layer (ABL) in a valley is often decoupled from the large-scale flow by a
strong temperature inversion . In this case mainly local
convection and thermally driven wind systems, which are caused by
differential heating of adjacent air masses, such as slope and valley winds,
determine the valley ABL e.g.. They
influence the diurnal temperature and humidity cycle
e.g. and also determine aerosol
dispersion and, thus, air quality e.g.. Mesoscale processes, such as topographic and advective venting,
accompany the thermally driven flows under fair weather conditions
e.g.. When the large-scale flow is not
negligible it also impacts the valley ABL. The interaction takes place via
turbulent transport or dynamically driven flow phenomena, which occur when
the large-scale flow is affected by the orography. Gravity waves, wave
breaking, downslope windstorms, hydraulic jumps and rotors can occur on the
lee side of the mountains. The intensity and extent of the developing
phenomena depend on the shape of the mountains, the stratification of the
atmosphere, the strength of the valley ABL inversion, the wind speed, and the
direction of the large-scale flow . Stratified flow
theory as well as hydraulic flow theory were both used successfully to
explain the aforementioned phenomena e.g.. Field campaigns were performed to
gather observations of these phenomena along large mountain barriers (Alps,
Pyrenees) during field campaigns such as ALPEX , PYREX
, and MAP and also for individual
valleys, e.g. T-REX , MATERHORN , or
craters like in METCRAX . This paper
contributes to this research by investigating a frequently occurring
mesoscale flow phenomena which influences weather and climate in the Dead Sea
(DS) valley.
The Dead Sea with a water level of currently -430 m above mean sea level
(a.m.s.l.) forms the lowest part of the Jordan Rift Valley, which is an over
700 km long north–south oriented depression zone extending from the northern
Israeli border to the Gulf of Aqaba. The complex and steep orography,
together with the land surface heterogeneity in the valley, introduced by the
lake, results in a strong local forcing and triggers pronounced thermally
driven wind systems, such as a lake breeze, slope, and valley winds
. Additionally, regional forcing influences the
atmospheric processes in the valley, resulting in a distinct diurnal wind
pattern. In particular, strong westerly downslope winds are observed
frequently in the evening. first described the westerly
winds in the northern part of the DS as the Mediterranean Sea breeze (MSB)
entering the valley, but not with the typical characteristics of a sea breeze
at the coast, i.e. a steady, cool, and moist air flow; they rather described it as
a very dry, hot, and gusty wind. Following various
observational near-surface studies
e.g. and also numerical
studies e.g. have been
carried out to study the penetration of the MSB into the Jordan Rift Valley.
Studies showed that the downward penetration of the MSB results from the
temperature difference between the cooler maritime air mass and the warmer
valley air mass. A density-driven flow, which accelerates and warms while
descending into the valley , i.e. a foehn wind, when
following the definition of the WMO . At the DS they occur
most frequently in summer and enter the Jordan Rift Valley first in the north
around Lake Kinneret and at approximately 18:00 LT (LT = UTC+2) at the DS
. These foehn events have a
large impact on the atmospheric conditions in the DS valley. Mean hourly wind
velocities of 5 m s-1 in the south and up to 12 m s-1 in the
north were recorded during these events . Furthermore, through the adiabatic heating of the air mass during
the descent into the valley, the foehn warms the valley . It also influences humidity. Some studies suggest an
increase in humidity , whereas others have suggested a
decrease . The foehn events influence the lake evaporation,
as evaporation is driven by wind velocity and vapour pressure deficit, as
shown by . The diurnal maximum of evaporation is reached,
untypically, shortly after sunset when the foehn sets in .
Finally, the foehn events can also cause an air mass exchange and remove the
aerosol particles and the often-occurring haze layer from the valley,
improving air quality .
Even though these foehn events apparently have such a large influence on the
atmospheric conditions at the DS, a detailed analysis of their
three-dimensional structure and their characteristics, such as height,
duration, and intensity, and further insights regarding their evolution are
missing. Hence, the following questions are addressed in this study. (i) Is a
differentiation between radiation and density-driven downslope flows
possible? (ii) What are the typical characteristics of the foehn events?
(iii) What are the key mechanisms during the foehn evolution?
For the analysis the first sophisticated high-resolution lidar measurements
of foehn events in the DS valley, along with long-term near-surface
observations, were used. The measurements were performed in the framework of
the interdisciplinary virtual institute DEad SEa Research VEnue (DESERVE)
. In the following section, Sect. , a
short geographical overview, information on the measurement sites and the
instrumentation as well as the applied methods is presented.
Section presents results of an objective occurrence
frequency analysis, characteristics of the foehn events are shown in
Sect. , and a detailed case study of a strong foehn event and
the processes leading to it are presented in Sect. .
Section provides a summary, including a conceptual model
of the processes, and conclusions.
Topographic map of the research area with the measurement sites (a). Cross section along the
dashed line in (a) is shown in (b), and a picture of the KITcube measurement site near Masada is shown in (c).
Methodology
Study area
The DS is the lowest reachable place on earth with a current water level of
-430 m a.m.s.l. The valley is north–south oriented, with the Judean Mountains
to the west, with a mean ridge height of about 895 m a.m.s.l., and the Moab
Mountains to the east, which reach up to 1200 m a.m.s.l.
(Fig. a). The cross section in Fig. b
illustrates the steep orography at both sides of the DS, extending over
1600 m in the vertical. The DS is about 100 km long and 15 to 17 km wide.
The distance to the Mediterranean Sea is 80 km in the north and increases up
to 120 km towards the south (Fig. a, b).
Experimental set-up
A long-term meteorological observation network has been operated around the Dead
Sea since 2014 (Fig. ). It was complemented by special
observation periods in August 2014 and November 2014, where the mobile
observatory “KITcube”, an assembly of ground-based in situ and remote sensing
instruments for probing the atmosphere , was deployed in
the DS valley, near Masada (Fig. a, c). Additional
radiosoundings were performed at the eastern shore of the DS in Ghor Haditha,
Jordan (Fig. a). Further details about the measurement
network around the DS and the KITcube instruments can be found in
and .
Instrumentation and data
The long-term measurements were performed using meteorological towers which
were equipped with sensors measuring air temperature, humidity, and up- and downwelling shortwave and longwave radiation components at 2 m above
ground level (a.g.l.), air pressure, and precipitation at 1 m a.g.l., with a
temporal resolution of 10 min. Supplementary measurements with a temporal
resolution of 20 Hz were performed with a sonic anemometer at 10 m a.g.l. at
the meteorological stations (Fig. a, orange stations) and
with an integrated gas analyser and sonic anemometer (IRGASON) at 6 m a.g.l. at
the energy balance stations (Fig. a, red stations),
providing all three wind components, sonic temperature, and water vapour.
Furthermore, long-term observations of wind, temperature, and pressure
in Jerusalem provided by the Israel Meteorological Service (IMS) and the
operational radiosoundings from the IMS at Bet Dagan launched everyday at
00:00 and 12:00 UTC were used.
The following subset of KITcube instruments was used in this study. The first instrument, a two-axis scanning pulsed 2 µm
wind lidar from Lockheed Martin called “WindTracer”, measured aerosol backscatter and radial velocity. It has a peak power of
4.5 kW and a pulse repetition frequency of 500 Hz .
The effective pulse length and hence the minimum spatial resolution is 56 m.
The scan pattern of the instrument included range-height-indicator (RHI)
scans at azimuth angles of 15, 62, 172, and 299∘ with an
elevation ranging from 0 to 180∘, and plan-position-indicator
(PPI) scans of 360∘ azimuth for elevation angles of 0.2, 5, 15,
and 75∘. The minimum detection range of the instrument is 400 m.
A quality threshold of -7 dB was used for the signal-to-noise ratio (SNR).
To close the gap between the surface and the minimum detection range of the
Windtracer, a second wind lidar from Leosphere was used. The “Windcube”
has a resolution of 20 m and a detection range from 40 to 600 m. It
operates with a wavelength of 1.54 µm and works with a 4-point
stop-and-stare mode at a fixed elevation angle of 75.2∘ and an
integration time of 7 s. From both lidar instruments, the vertical profiles
of the horizontal wind were calculated using the velocity–azimuth display algorithm after
. Radiosondes were simultaneously launched during
intensive observation periods on both sides of the DS every 2 h, always from
Friday 11:00 UTC until Sunday 03:00 UTC due to airspace restrictions,
providing temperature, humidity, wind velocity, and wind direction profiles.
Layer detection algorithm
The characteristics of the foehn events were derived from lidar
RHI scans along the main wind direction of the foehn
events (black dashed line in Fig. a). By introducing
a layer detection algorithm the characteristics could be derived objectively.
The algorithm uses profiles of the horizontal component of the radial
velocity (Fig. b) and the backscatter signal
(Fig. c), averaged between 2 and 3 km away from the
lidar. The profiles show that within the foehn layer the flow has a jet-like
structure with strong wind velocity and backscatter gradients
(Fig. d). The strongest wind velocity and backscatter
gradient above the wind velocity maximum and the backscatter minimum,
respectively, were used within the algorithm to determine the foehn layer
height (red dashed line in Fig. d). Furthermore, the
algorithm detected the mean and maximum wind speed, and the height of the
wind speed maximum from each RHI scan.
These results were then used to derive averaged characteristics for each event.
PPI scan of the radial velocity at 5∘ elevation on 16 August 2014, 19:01 LT (a).
Blue colours indicate flow towards the lidar and red colours flow away from the lidar. The dashed
line indicates the direction of the RHI scans of the horizontal component of the radial velocity shown
in (b) and of the backscatter signal shown in (c) (note the different colour scales). RHI scans were
performed on 16 August 2014, 19:06 LT. Vertical profiles of the horizontal radial wind velocity component
and of the backscatter signal (d) averaged between the black dashed lines in (c). The red dashed lines
indicate the foehn layer heights, automatically detected by the strongest vertical gradient of the
respective profile above the maximum.
Reduced-gravity shallow-water theory
To explain the observed phenomena described in this paper, the
reduced-gravity shallow water (RGSW) theory, as described by
, is applied. It is a theory to describe dynamically driven
flows over mountains. The RGSW theory describes a two-layer system separated by a
strong temperature inversion. An air mass can either be blocked and forced to
flow around the mountain, it can flow over the mountain, or it can pass
through gaps in the mountain range. The ratio between the energy needed to
overcome negative buoyancy and the horizontal inertia of the air mass
is called the non-dimensional mountain height,
H^=NhU,
with the crest height, h, the horizontal wind speed upstream, U, and the
Brunt–Väisälä frequency, N. H^ is normally calculated for a layer,
and thus the values represent mean values of this layer. For low wind speeds
or strongly stable stratification the air mass is blocked by the mountain
indicated by H^>1, but for H^≤1 the flow has sufficient
kinetic energy to flow over the mountain. To explain the characteristics of
the flow over the mountain the Froude number,
Fr=ug′η,
where u is the mean wind velocity of the lower layer, η is the inversion
height, and g′=g⋅ΔΘ/Θ, the reduced gravitational
acceleration, can be used. ΔΘ is hereby the inversion strength
at the top of the lower layer and Θ the mean potential temperature of
the lower layer. The flow is subcritical if Fr<1, Fr>1 means the flow is
supercritical, and Fr=1 is called the critical state. Three situations can
occur when a flow passes an obstacle. Firstly, the layer is subcritical
everywhere. Then the layer height has to decrease when the terrain rises and
it has the shallowest depth at the crest. Secondly, the flow is supercritical
everywhere; then the layer height and the terrain height rise simultaneously.
Thirdly, the layer changes from a subcritical state upstream to supercritical
state downstream. This means that the layer has to be thick and slow upstream
and fast and shallow downstream. Further downstream the layer eventually
thickens and slows down in a hydraulic jump. This causes strong turbulence as
the momentum is conserved but kinetic energy dissipates .
This is one of the oldest conceptual models and was first proposed by
, who suggested that there is a fundamental similarity
between downslope wind storms and hydraulic jumps. For the transition of the
state from subcritical to supercritical, the flowing layer has to reach its
critical state Fr=1 at crest height.
Results
Occurrence frequency
To analyse how often the westerly downslope winds are actually
density-driven, the occurrence frequency of foehn was calculated using the
automatic and probabilistic statistical mixture model of
with data from the long-term observations between 2014 and 2016. This model
enables a probabilistic distinction between the density-driven foehn events
and other downslope winds, such as nocturnal radiatively driven downslope
flows. Density-driven flows are possible when the potential temperature of
the upstream air mass crossing at crest height is equal to or lower than the
temperature in the valley. The air mass descends, resulting in similar
potential temperatures at the crest and in the valley. In contrast,
radiatively driven downslope flows are triggered by radiative cooling of the
air layer near the slope and the resulting temperature gradient between the
air at the slope and the air at the same height in the valley centre
. This leads to a stable stratification and therefore to
a positive potential temperature difference between the crest and the valley
(ΔΘ=Θcrest-Θvalley>0). The model is based on the
assumption that the probability density function of these two wind regimes
are statistically distinguishable in one or more characteristic variables.
The model fits parametric distributions for the two wind regimes whose
mixture results in the observed probability distribution of all downslope
wind cases. It yields the probabilities that measurements at a particular
time are density-driven or radiatively driven downslope flows (for further
details see ). The wind regimes are identified using the
potential temperature difference (ΔΘ) between the crest
(Jerusalem, 810 m a.m.s.l.; Fig. a) and two downstream
stations, one at the slope (Masada, -7 m a.m.s.l.; Fig. a)
and one in the valley (Ein Gedi Beach, -427 m a.m.s.l.;
Fig. a), as well as the wind speed at the respective
downstream station. The only parameter which has to be set prior to the fully
automatic classification is the wind direction sector indicating
“downslope”. For Masada, this is 200–315∘ and for Ein Gedi
Beach 220–320∘.
(a, b) Probability of foehn (colours) for wind velocity and potential temperature difference
(ΔΘ=Θcrest-Θvalley). The crest station is Jerusalem and the valley
stations are Masada and Ein Gedi (Fig. ). (c, d) Relative frequency of foehn
occurrence with a detection probability of more than 75 %. Sunrise and sunset are marked as red lines.
Figure a and b show the likelihood of foehn for the
potential temperature difference and the wind velocity in the valley. The two
wind regimes can easily be distinguished. During nocturnal radiatively driven
downslope flows the radiative cooling leads to a stable stratification of the
valley ABL, resulting in high positive temperature differences of up to 13 K
in the valley and 7 K at the slope, and wind velocities between
1–3 m s-1. This coincides with literature values of 1–5 m s-1
for the maximum wind velocities of radiatively driven downslope flows
. During foehn events, the air mass at crest
height descends into the valley, leading to a smaller temperature gradient
between -4 to 2 K at the slopes and -3 to 2 K in the valley. Wind
velocities accelerate down the slope and can reach up to 16 m s-1. The
climatological distribution of foehn (Fig. c, d) shows
that it occurs most often in the afternoon in summer. They do not occur
before 15:00 LT at the slopes and 16:00 LT in the valley. The occurrence
frequency is highest directly after sunset with 72 % at the slopes and
still 64 % in the valley. In the valley the foehn generally ceases around
midnight, whereas at the slope foehn is still observed in up to 16 % of the
nights. In summer it ceases shortly after sunrise, whereas in winter foehn
might continue after sunrise.
Thus, on a diurnal as well as on an intra-annual basis foehn is observed more
frequently at the slope than in the valley.
There are two mechanisms which could be responsible for these differences.
Firstly, the foehn detaches from the slope, as it reaches the level of
neutral buoyancy in the valley and continues flowing in an elevated layer
above the valley floor . Secondly, strong along-valley flows
dominate and disturb the penetration of foehn into the valley. However, with
the long-term near-surface observations alone a further investigation of
these differences is not possible.
Characteristics of foehn in the Dead Sea valley
To get detailed information about the three-dimensional characteristics of
the foehn events, lidar data from the special observation period in August 2014 were analysed. Figure shows the occurrence frequency
of foehn for August 2014 as detected by the probabilistic statistical mixture
model. On 26 days foehn was detected at the slope (blue dots), whereas in the
valley only 22 events (red dots) were detected. For 13 events data from the
scanning lidar were available.
Occurrence of foehn in August 2014 at the slope (Masada, blue dots) and in the valley
(Ein Gedi, red dots). Shown is the foehn occurrence with a detection probability of more than
75 %. Grey shaded area shows the dates when lidar data were not available.
Characteristics of the foehn events derived from RHI scans at 299∘
(dashed line in Fig. a). Presented are mean wind speed, mean height, mean maximum wind
speed, and mean height of the wind speed maximum of the foehn events. The mean wind speed was first calculated
for each RHI scan and then averaged over the whole event. The height, maximum wind speed, and height of the wind
speed maximum were determined for each RHI scan and then also averaged over the whole duration of the event. The
events are sorted according to their duration.
Date
Mean wind speed
Mean layer height
Mean maximum
Mean height of
Duration
Observed
(m s-1)
(m a.g.l.)
wind speed
maximum wind speed
(hh:mm)
at KITcube
(m s-1)
(m a.g.l.)
location
13.08.2014
3.4
420.0
5.1
199.0
01:51
yes
22.08.2014
4.1
529.1
5.6
293.6
02:09
yes
25.08.2014
6.2
501.8
8.3
312.7
02:39
no
14.08.2014
3.0
600.0
3.1
407.4
03:14
no
26.08.2014
7.2
961.1
10.8
406.8
04:13
yes
15.08.2014
7.5
1338.5
10.1
574.8
05:12
yes
16.08.2014
6.8
924.7
8.7
493.0
05:18
yes
18.08.2014
5.1
834.4
7.5
450.9
05:24
yes
28.08.2014
5.4
386.5
8.3
201.8
11:47
yes
31.08.2014
4.0
251.0
6.1
124.0
11:41
yes
For two events (19 and 27 August), which were detected by the probabilistic
model, the scanning lidar was not working during the foehn event; however,
the Windcube, covering the lowest 40–600 m a.g.l., was working and showed a
strong westerly flow at the measurement location, confirming the foehn
occurrence. The 13 measured events give us the opportunity to derive further
characteristics of the foehn and additionally check the aforementioned
hypothesis of elevated foehn layers. For 10 of the 13 events the
characteristics could be derived using the layer detection algorithm
(Sect. ). Manual inspection of the data for
the three other events (17, 24, and 29 August) shows that on these three days
the algorithm did not detect the layer characteristics because the foehn was
only observed in an elevated layer above the ground at the KITcube
measurement site. Only for 29 August were radiosondes available in the
afternoon, showing a temperature inversion at -220 m a.m.s.l. and a westerly flow
just above this inversion. This supports our hypothesis that the foehn
reached the level of neutral buoyancy at an elevated layer above the valley
floor (not shown).
The characteristics of the other 10 events, derived with
the described algorithm, are shown in Table . The
results of the analysis show considerable differences between the events
regarding their duration, strength, and vertical extent. They could be grouped
into two main categories: (i) the weak, shallow, and short events, and (ii) the
longer-lasting, strong events with a large vertical extent. Two events
did not fit in one of the groups as they had a very long duration of over
11 h and varying layer characteristics over the course of the event;
therefore they are called “mixed events”. An example for the two main types is shown
in Fig. .
Example for a short and weak foehn event (a, b, c) and for a strong foehn event (d, e, f).
Shown are 10 min mean vertical profiles of horizontal wind at the KITcube location (a, d). Length
and colours of the arrows represent wind velocity and the direction of the arrows indicate the
horizontal wind direction.
Time–height cross sections of the averaged vertical profiles of the horizontal radial wind
component derived from lidar RHI scans at 299∘ (b, e). Negative values (blue colours)
indicate a wind component from north-west and positive values (red colours) indicate a wind
component from south-east. Black stars indicate the detected height of the foehn. Below the
cross sections the variance of the vertical wind (σw) at 2 m a.g.l. and wind direction
(wd) measurements at 40 m height are shown. Hovmoeller diagrams of the near-surface radial
velocity from lidar RHI scans at an azimuth angle of 299 and 7∘ elevation west
of the lidar and 1.6∘ east of the lidar (c, f). The black dashed line indicates the end
of the western slope and the beginning of the valley floor. Wind direction (wd) was measured
at 40 m a.g.l. Orange points highlight wind directions between 270 and 330∘, the main wind direction of the foehn.
Weak, shallow, and short events
The first type of foehn is a short and weak foehn wind, which only partly
penetrates into the valley. The example from 22 August 2014 lasted about
2 h from 18:00 to 20:00 LT and showed maximum near-surface wind velocities of
about 5 m s-1 (Fig. a). Before the west wind
reached the valley, an elevated west wind maximum just above crest height
(1750 m a.g.l.) was observed in the afternoon. At 16:00 LT, it started to
penetrate down into the valley and reached the valley floor at around
18:30 LT (Fig. b). To evaluate how far the foehn
penetrated into the valley Hovmoeller diagrams were derived from lidar RHI
scans along the main flow axis of the foehn using the lowest elevation angle
above the orography.
The Hovmoeller plot in Fig. c shows how the foehn
penetrated down the slopes, starting at around 16:00 LT. An opposing upslope
wind was still present during this time, slowing down the foehn. At around
18:00 LT the upslope wind ceased and the foehn penetrated down into the
valley. For a time window of about 1 h the foehn reached approximately 6 km
into the valley and was also observable at the KITcube location, which was
3.5 km south-east of the slope (Fig. c). The mean wind
speed of the foehn was 4.1 m s-1 and the mean layer height 529 m
(Table ). The mean maximum wind velocity was detected
at around 300 m a.g.l. with a mean wind velocity of 5.6 m s-1.
The
characteristics of the other three events of this type are summarised in Table . They all show very similar characteristics. A rather
short duration of 2 to 3 h, mean wind velocities of 3 to
4 m s-1 (except for one event with 6.2 m s-1), a mean layer
height of 400 to 600 m, and the mean maximum velocity in a height of 200 to
400 m. Two of the four events reached the valley floor and penetrated
approximately 6 km into the valley, whereas the other two foehn events were
only observed close to the slope and in an elevated layer above the KITcube
location (Table ).
Longer-lasting and strong events
The second example is the 16 August 2014, representing a very strong event
with wind velocities exceeding 10 m s-1 (Fig. d).
Similar to the first case, an elevated west wind maximum around crest height
penetrated into the valley, but it already started at 14:00 LT and affected
a deeper layer of about 1500 m (Fig. e). It was
observed in the valley at 16:30 LT. Another difference compared to the first
type is the strong wind velocity increase while the foehn descended into the
valley (Fig. d). At crest height velocities of about
6 m s-1 increased to 9 m s-1 in the valley and the layer height
decreased over the course of the event from 1500 to only 350 m. This
strong foehn also penetrated further into the valley. It was observed at the
KITcube location for nearly 4 h (Fig. f). The
Hovmoeller plot shows that the foehn penetrated at least 13.5 km into the
valley, with wind velocities up to 10 m s-1. It could not be
determined how far it actually reached into the valley, as the lidar
measurements were limited to a radius of 10 km (Fig. f).
However, radiosondes launched at the eastern shore of the DS
(Fig. ) indicate that the foehn reached over the DS
towards the eastern shore. Soundings at 17:00 and 19:00 LT also showed
north-westerly winds advecting a drier air mass towards the eastern shore,
although wind velocities were weaker than on the western side. The general
characteristics of the other strong events were similar to the described one.
The overall duration of the strong events varied between 04:00 and 05:30 h,
the mean wind velocity of the foehn was 5 to 7 m s-1, and mean maximum
wind velocity was 7.5 to 11 m s-1 at a height of about 400 to 570 m.
The penetration distance of the four events all exceeded the measurement
range of the lidar, meaning that they penetrated at least 13.5 km into the
valley. For the 15 and 16 August radiosondes were launched at the eastern
shore, confirming that the foehn reached the other side of the valley.
Mixed events
The two mixed events had a duration of over 11 h. The other characteristics
were similar to the weak events, with mean wind velocities of 4 and
5.4 m s-1; the mean layer height was 251 and 387 m and the mean
maximum wind velocity was 6.1 and 8.3 m s-1, respectively. However,
the Hovmoeller plot shows that in particular on the 28 August the event can
be divided into two phases. First, a strong downslope wind with velocities of
10 m s-1 persisted for about 4 h between 18:00 and 22:00 LT and then
transformed into a shallow layer with wind speed around 5 m s-1
(Fig. ). For the first 4 h, the foehn penetrated over
13 km into the valley, but in the second phase it merely reached the lidar
location. The observations suggest that on 28 August a strong downslope wind
event with strong wind speeds transformed into a shallow downslope flow
lasting for the rest of the night.
Example for a mixed event on 28 August 2014. Hovmoeller diagram of the near-surface radial velocity
from lidar RHI scans at an azimuth angle of 299 and 7∘ elevation west of the lidar and
1.6∘ east of the lidar. The black dashed line indicates the end of the western slope and the
beginning of the valley floor. Wind direction (wd) was measured in 40 m a.g.l. Orange points highlight
wind directions between 270 and 330∘, the main wind direction of the foehn.
Process understanding of foehn events affecting the whole valley
The foehn events which penetrate far into the valley obviously have the
strongest effects on the valley atmosphere. Through the deeper layer and the
further penetration into the valley, they cause an air mass exchange,
changing temperature, humidity, and aerosol concentration as well as
increasing evaporation from the lake surface. The processes which cause such strong
events can not be determined by the results presented so far. Therefore, a
case study was selected and a detailed analysis of the atmospheric conditions
was performed to reveal the relevant processes leading to such strong events.
As a case study the event of the 16 August 2014 was selected. On 16 August a
typical summer synoptic system, a shallow Persian trough, extended from the
Persian Gulf over Iraq and Syria bending north-west towards the Mediterranean
Sea and Greece (Fig. ). This resulted in westerly
flow in the lowest 1000 m a.m.s.l. downstream of the trough axis. Offshore over
the Mediterranean Sea the near-surface winds had an intensity of about
4 m s-1 in the morning. The intensity of the trough strengthened over
the course of the day and the near-surface westerly flow over the
Mediterranean Sea increased to over 7 m s-1 in the afternoon.
Large-scale synoptic conditions on 16 August 2014, 15:00 UTC. Geopotential in gpdm at 500 hPa
(black contours), surface pressure in hPa (white contours), relative topography 500–1000 hPa (coloured).
Green dot indicates investigation area.
The evolution of the ABL can be divided into three stages.
Stage I: ABL evolution prior to the foehn (07:00–15:00 LT)
In the morning a strong temperature inversion marked the height of the
valley ABL at 900 m a.m.s.l., resulting in a well-defined capping stable layer
(Fig. a). In the valley, two local
thermally driven wind systems developed. Upslope winds set in around
07:00 LT,
as indicated by the measurements at the slope (Masada) shown in
Fig. . The wind direction changed from west to
east, and the vertical wind velocity changed sign from -0.4 to
0.4 m s-1. The easterly DS lake breeze reached the KITcube location at
09:00 LT. Wind direction turned east, the 2 m wind velocity increased to
2 m s-1, and the diurnal temperature increase over land was attenuated
by the advection of the cooler air from the water until 16:30 LT
(Fig. ). The onshore flow of the lake breeze
reached up to 350 m a.m.s.l. with a mean wind speed of 2.4 m s-1, and the
return flow was observable between 500 and 900 m a.m.s.l. in the radiosonde
profile at 09:00 LT (Fig. b). Above the
temperature inversion, the large-scale flow was from the north-west with a wind
velocity of about 11 m s-1. The strong vertical wind shear between the
easterly lake breeze and the strong westerly large-scale flow caused
mechanically induced turbulence (Ri<0.25), which led to a downward mixing of
warmer air into the layer between 350 and 900 m a.m.s.l. between 09:00 and
13:00 LT, which also increased potential temperature
(Fig. a). A temperature inversion formed at
around 550 m a.m.s.l., resulting in a stable layer with a potential temperature
increase of 2 K. A secondary weaker inversion at 1200 m a.m.s.l. represented
the former ABL top at 13:00 LT. Between the two inversions a layer with
westerly flow and reduced wind velocities compared to the large-scale flow
was established (Fig. b). To understand the
entire development in the valley the upstream conditions are also relevant.
The analysis of the numerical weather forecast model COSMO-EU
showed the development of a near-neutral convective
boundary layer (CBL) over land in the morning. Through the westerly
large-scale flow, caused by the shallow Persian trough, the evolution of a
well-defined Mediterranean Sea breeze front could not be observed
(Figs. a, a). Neither
a temperature decrease nor a humidity increase, the typical characteristics
associated with a sea breeze front propagating inland, were observable over
the coastal plains. Also, in the wind field a possible MSB can not be
distinguished from the large-scale flow. The advected air masses were mixed
within the CBL and no frontal structure of the MSB could be observed. In the
afternoon at 14:00 LT the large-scale near-surface flow over the
Mediterranean Sea strengthened to approximately 8 m s-1 and advected
more stratified moist and cool air towards the coastal plains
(Fig. b). The inversion layer inhibited the CBL
evolution. Potential temperature in the CBL over the plains stagnated at
around 302 K and the thickness of the CBL decreased from around 960 m at
11:00 LT (Fig. a) to 620 m at 14:00 LT
(Fig. b). At the same time at the mountain ridge
the potential temperature of the CBL increased by 3 K and CBL height
increased from 735 to 910 m a.g.l. This indicates that the maritime
air mass did not reach up the slopes at that time
(Fig. b).
Potential temperature (a) and horizontal wind profiles (b), measured with radiosondes in
the valley at the KITcube location. Data are shown for 16 August 2014.
Time series of potential temperature (Θ), potential temperature difference between crest and
slope/valley (ΔΘ), wind direction (wd), wind speed (ws), vertical wind speed (w),
specific humidity (q), and turbulent kinetic energy (TKE) for Jerusalem (crest), Masada (slope), and KITcube location (valley) on 16 August 2014.
Stage II: development of the foehn (15:00–18:30 LT)
In the afternoon the temperature gradient between the air at the mountain
ridge and in the valley changed sign at 15:00 LT
(Fig. ). At the same time wind direction at the
slope changed from east to west, horizontal wind velocity suddenly increased,
vertical wind velocity became negative, and specific humidity dropped
(Fig. ). This indicates that the air mass at ridge
height started to descend into the valley. The descent and acceleration of
the air mass can also be observed in the model results
(Fig. b). In the valley itself, the foehn was
observed at the KITcube location at 16:30 LT with a near-surface wind
velocity of 5.5 m s-1 and a strong increase in turbulent kinetic energy (TKE) to
3 m2 s-2 (Fig. ). The vertical extent of
the foehn, calculated from lidar data, was 1500 m a.g.l., with a mean wind
velocity of 6.0 to 7.0 m s-1 (Fig. e). The
temperature profile of the radiosonde at 17:00 LT showed a near-neutral
layer up to 1250 m a.g.l. with 308 K, which was the temperature of the
elevated layer between 550 and 1100 m a.m.s.l. at 15:00 LT
(Fig. a). This is another indicator that
the air from the elevated layer descended into the valley and replaced the
local air mass. From 17:00 until 18:30 LT the density current had a layer
height of about 1400 to 1050 m a.g.l. and the horizontal radial
mean wind velocity was 6.0 to 6.5 m s-1 (Fig. e).
West–east cross section of specific humidity (coloured) and potential temperature
(isolines) at the latitude of Masada (31.3125∘ N) from COSMO-EU analysis data for
16 August 2014.
West–east cross section of u-wind component (coloured) and potential temperature
(isolines) at the latitude of Masada (31.3125∘ N) from COSMO-EU analysis data for 16 August 2014.
Stage III: intensification of the foehn (18:30–21:00 LT)
The COSMO-EU analysis showed that upstream, with decreasing solar radiation
in the evening, convection weakened and potential temperature decreased. The
mean potential temperature in the CBL over the coastal plains was 300.6 K,
and at ridge height it was 304.3 K at 17:00 LT
(Fig. c). A comparison of the near-surface model
results with station observations from the IMS at Jerusalem and Bet Dagan
showed comparatively good agreement; however, the temperature decrease in the
afternoon occurred too early at both stations in the model. A time shift of
1.5 h was observed, which means that the upstream afternoon conditions
described from the model results have to be shifted from 17:00 to
18:30 LT. At that time model results show that the potential temperature at
the CBL top increased by 4.1 K over the coastal plains and at the mountain
ridge by 5.0 K. The mean wind speed within the ABL increased to
5.1 m s-1 near the coast and 7.7 m s-1 at the ridge and the CBL
height decreased considerably to 360 m a.g.l. over the plains and 330 m a.g.l.
over the mountain ridge (Fig. c). Measurements
in the valley and at the slope show that the mean wind velocity of the foehn
increased to about 9 m s-1 at 18:30 LT, and the height of the foehn
decreased to 350 m a.g.l. (Fig. e). Also, the near-surface
measurements showed an increase in the wind velocity and turbulent kinetic
energy (Fig. ). The radiosonde profiles at
19:00 LT showed a strong stable layer with a potential temperature increase
of 3.3 K at around 270 m a.g.l. (-80 m a.m.s.l.) and the potential temperature
in the layer was 305.6 K (Fig. a). The
profile of the horizontal wind had a clear jet-like structure
(Fig. b). Around 19:00 LT a sudden increase
in the layer height and a reverse flow near the surface were detected east of
the lidar (Fig. a). There, the height of the foehn
layer increased to approximately 1000 m a.m.s.l. and the air below was quite
turbulent (see also animation of lidar measurements in the Supplement). The
backscatter RHI scan shows that the air mass in the reverse flow has the same
aerosol content as the air mass of the foehn layer, indicated by the same
backscatter values of -7.7 dB. The air mass above the foehn layer has much
lower backscatter values of -8.25 dB (Fig. b). The
high backscatter value of the foehn layer most likely results from local near-surface dust emissions, caused by the high wind velocities and the increased
turbulence.
The sudden increase in the layer height together with the near-surface
reverse flow and increased turbulence indicates a hydraulic jump with a rotor
formation beneath. This phenomenon can be explained by using the
reduced-gravity shallow-water theory. It results when a subcritical flow
upstream changes to a supercritical flow downstream. Then a sudden transition
of the supercritical flow back to a subcritical state occurs by converting
some of the flow's kinetic energy into an increase in potential energy, i.e.
a hydraulic jump. Calculating the Froude numbers for the coastal plain (from
model), mountain ridge (from model), and valley (from measurements) results
in Frplains=0.73, Frridge=1.06, and Frvalley=1.7, which
confirms the assumption of a hydraulic jump in the valley. The reverse flow
of the rotor causes the formation of a convergence line near the surface
(Fig. c). However, this convergence line is not
symmetrically aligned with the mountains. North of the lidar, the convergence
line is located close to the shoreline of the DS, whereas south-east of the
lidar location it is much closer to the measurement site. The non-uniform
distance from the mountains results most likely from the specific local
orographic characteristics west of the lidar. North-west of the lidar a wadi
opens into the Dead Sea valley. The westerly flow is most likely channelled
in the wadi, resulting in accelerated wind speeds, and thus a stronger
penetration into the valley, locating the convergence line further east. The
convergence line was not stationary but retreated towards the western slope,
whereby the rotor grew horizontally and vertically over the course of the
event. On 21:00 LT the convergence line was observed at the KITcube
location, with maximum TKE values of 4.6 m2 s-2. As it
progressed further towards the slope the TKE and wind velocity dropped at the KITcube
location (Fig. ). The foehn ceased around
21:30 LT, caused by further stratification of the upstream air masses, which
led to a blocking of the flow at the Judean mountains, indicated by a
non-dimensional mountain height of about 2.7 at 20:00 LT in the model,
meaning about 21:30 LT in reality.
RHI scan at an azimuth angle of 299∘ (dashed line in c) showing radial velocity (a)
and backscatter (b) on 16 August 2014, 19:36 LT.
PPI scan at an elevation angle of 0.2∘, shows radial velocity (c) at 19:51 LT. In (a) and
(c) blue colours indicate a flow towards the lidar and red colours indicate a flow away from the
lidar. The shoreline of the Dead Sea is indicated by a dark blue line in (c).
Conceptual model describing the different boundary layer processes in the
valley on 16 August 2014. (a) The temporal evolution of the different layers and
wind systems. The coloured areas refer to the distinct layers detected. Hatched areas
describe distinct processes described in the text, and red arrows in the legend show
the main wind direction in the respective layer. Panel (b) shows the wind systems (arrows)
and the ongoing processes (shaded areas) leading to the next stage for morning, noon, afternoon, and evening.
Summary and conclusion
This study investigated frequently occurring foehn events in the DS valley
using long-term near-surface observations from 2014 until 2016, as well as
high-resolution lidar measurements which were performed from these foehn
events for the first time in 2014. With the automatic and probabilistic
statistical mixture model from an objective
identification and separation of foehn from the radiatively driven downslope
flows was possible. The results confirmed the findings from earlier studies
e.g. that throughout the year,
foehn evolves in the evening around sunset. The new results reveal a higher
occurrence frequency at the slopes than at the valley floor, indicating that
the foehn penetrates downslope most of the days, but does not reach the
valley floor on all occasions. We assume that the foehn detaches from the
slope at the level of neutral buoyancy in the valley and continues flowing in
an elevated layer above the valley floor. In the lidar data these elevated
layers were observable and in one case radiosonde ascents were also
available, which showed a low-level inversion in the valley, strengthening
our assumptions. These events are similar to cases in the Owens valley
described by , where gap flows also detached from the slope
at the level of neutral buoyancy.
Two main types of foehn in the DS valley could be identified with these new
data: weak and rather short events and strong, longer-lasting events which
affect the whole valley, leading to a complete air mass exchange. In contrast
to major downslope windstorms in other regions of the world
e.g. foehn at the
DS occurs on a diurnal scale and only for a few hours per evening. This
already shows that it is not synoptically driven but depends on diurnal local
and mesoscale processes, such as the delayed development and cooling of the
valley ABL or the MSB reaching the Judean mountains in the afternoon, both
leading to a horizontal temperature gradient across the mountain range. In
the presented case of a strong foehn event, the delayed cooling of the valley
ABL in the afternoon led to the initiation of the foehn, which was then
intensified by supercritical flow conditions, leading to a hydraulic jump and
a rotor formation in the valley. A conceptual model was developed to
summarise and illustrate the temporal and spatial interactions of the
responsible mechanisms for such an event, based on the results of the 16 August 2014 (Fig. ). The model was divided into four steps:
Morning. A strong temperature inversion decoupled the valley ABL
from the large-scale flow and thermally driven upslope winds and a lake
breeze developed in the valley after sunrise (Fig. ). A
strong wind shear between the easterly lake breeze in the valley and the
strong westerly large-scale flow produced turbulence at the top of the valley
ABL, indicated by a Richardson number below 0.25 (striped area in
Fig. a and b).
Due to the turbulence, warmer air from above was mixed into the upper part of
the valley ABL, and a secondary temperature inversion below mountain ridge
height evolved. Even though the turbulent mixing decreased the inversion
height, an undisturbed valley ABL existed below the inversion.
Midday. Between the two temperature inversions below and above ridge
height a westerly flow with reduced wind speed formed a layer between the
valley ABL with its easterly lake breeze and upslope winds and the
large-scale strong westerly flow. The valley itself was further heated by
radiative forcing, which was concentrated within the reduced air volume in
the valley below the lower inversion. The topographic amplification factor
strengthened the valley heating
additionally.
Afternoon. At the ridge radiative cooling set in and eroded the
inversion below ridge height. The cooler air from the flowing layer formed a
deep foehn flowing down the slope. While descending into the valley the wind
velocity accelerated. A change in the upstream conditions triggered the
further development. The upstream subcritical flow reached a critical stage
around ridge height, as indicated by the calculated Froude numbers of 0.73
over the coastal plains and 1.06 at the mountain ridge.
Evening. This transition upstream led to supercritical flow
conditions downstream in the valley (Fr=1.7). The layer height of the foehn
in the valley decreased strongly and wind velocity increased. In
the valley the flow went from supercritical to a subcritical state forming a
hydraulic jump with a rotor beneath. The layer height increased suddenly and
mean wind velocities decreased, generating severe turbulence. The rotor led to
the formation of a convergence line near the ground. Finally, blocking of the
upstream air masses by the mountains stopped the foehn, as indicated by a
non-dimensional mountain height of 2.7.
This study showed that foehn at the DS is initiated by the horizontal
variation in boundary-layer temperature across the mountain range. This is
caused by an amplified heating of the valley ABL together with a delayed ABL
development in the DS valley compared to the upstream boundary layer, as
presented in this study. It can also be caused by the arrival of the cool MSB
at the western side of the mountains, as shown by various other studies
. The obtained
results of this study are highly relevant for the DS region. They show that
the foehn events are variable with respect to their strength, layer height, and
penetration into the valley. This is particularly important for the
temperature, humidity, and aerosol distribution in the valley, as well as
for the evaporation from the DS water surface. In particular, the coupling to
evaporation is of high importance as evaporation is the main loss of water
from the DS , determining the rate of the DS shrinking. We
conclude that for correct climate, weather, and also evaporation forecasts in
the region it is therefore important to correctly represent the boundary
layer processes in the valley, in particular the inversion heights and the
diurnal valley heating, as well as the occurring thermal wind systems, the
ABL development upstream, and the interaction with the MSB. In light of
the ongoing climate change, and the shrinking of the DS, daily maximum
temperatures in the valley could further increase, making a penetration of
foehn down into the valley more likely in the future. This of course would
further increase evaporation and the shrinking of the DS, resulting in a
positive feedback cycle and changing the diurnal temperature and humidity
range.
These results also contribute to the wider understanding of the boundary layer
processes in smaller valleys under weak synoptic forcing, where thermally
driven local and mesoscale wind systems govern the wind field. Thermally
driven foehn events similar to the ones in the DS valley were also observed
in other areas of the world. The Washoe Zephyr occurs at the eastern slope of
the Sierra Nevada and is driven by the asymmetric heating between the lower
western side and the elevated semi-arid Great Basin . At the
lee side of the Cascade Mountains the winds are triggered by the temperature
gradient between the semi-desert area on the lee side and the marine air
masses on the upstream side close to the Pacific Ocean coast
and a similar phenomena was found by
for the Mexico City basin. Even though these wind systems differ in scale,
the trigger mechanism is similar in all cases and they can have a large
impact on the local aerosol concentration and dispersion as well as on the
local climatic conditions.