An Observational Study of the Mesoscale Mistral Dynamics

We investigate the mesoscale dynamics of the mistral through the wind profiler observations of the MAP (autumn 1999) and ESCOMPTE (summer 2001) field campaigns. We show that the mistral wind field can dramatically change on a time scale less than 3 hours. Transitions from a deep to a shallow mistral are often observed at any season when the lower layers are stable. The variability, mainly attributed in summer to the mistral/land–sea breeze interactions on a 10-km scale, is highlighted by observations from the wind profiler network set up during ESCOMPTE. The interpretations of the dynamical mistral structure are performed through comparisons with existing basic theories. The linear theory of R. B. Smith [Advances in Geophysics, Vol. 31, 1989, Academic Press, 1–41] and the shallow water theory [Schär, C. and Smith, R. B.: 1993a, J. Atmos. Sci. 50, 1373–1400] give some complementary explanations for the deep-to-shallow transition especially for the MAP mistral event. The wave breaking process induces a low-level jet (LLJ) downstream of the Alps that degenerates into a mountain wake, which in turn provokes the cessation of the mistral downstream of the Alps. Both theories indicate that the flow splits around the Alps and results in a persistent LLJ at the exit of the Rhône valley. The LLJ is strengthened by the channelling effect of the Rhône valley that is more efficient for north-easterly than northerly upstream winds despite the north–south valley axis. Summer moderate and weak mistral episodes are influenced by land– sea breezes and convection over land that induce a very complex interaction that cannot be accurately described by the previous theories.


Introduction
The mistral is a northerly, low-level, orography-induced, cold-air outbreak over the north-western Mediterranean. It accompanies cold and dry continental air masses and restores clear sky conditions after the passage of a cold front over Provence (south-eastern France). During summer, it accelerates ground drying and is responsible for the propagation of devastating forest fires (Wrathall, 1985). Furthermore, mistral gusts can cause severe damage to  Table II. Geostrophic winds given in Table I are computed from the ECMWF analysis along the transects AB and CD. The dashed line in grey indicates the Rhone valley axis.
obstacles (Scha¨r and Smith, 1993). From upstream conditions taken at Lyon (see Figure 1), Pettre´(1982) gives a description of mistral behaviour arguing that the horizontal distribution is closely linked with hydraulic jumps. In downslope windstorms, Smith (1985Smith ( , 1989 found that wave breaking is associated with hydraulic jumps and in turn associated with the generation of potential vorticity (Smolarkievicz and Rotunno, 1993;Scha¨r and Smith, 1993). Drobinski et al. (2001a) have shown that the location at which a hydraulic jump occurs depends on the upstream and downstream flow conditions and not only from upstream conditions. Recent high resolution simulations have shown that the western and eastern boundaries of the mistral are partly defined by gravity wave breaking over the Massif Central and the Alps, respectively (Jiang et al., 2003). Recently, P. Drobinski et al. (2004, private communication) have found that additional processes such as wall separation on the western flanks of the southern Alps are involved in the Alps wake formation. As suggested by Jiang et al. (2003), the Massif Central wake separates the mistral from the tramontane. The latter flow blows between the Massif Central and the Pyre´ne´es and is considered as the companion of the mistral since they have the same synoptic origin and often blow simultaneously (Georgelin and Richard, 1996;Drobinski et al., 2001b).
Although the large-scale features of the mistral are well described, mesoscale aspects such as the temporal, vertical and horizontal variability, its onset and cessation are still to be investigated. During MAP (Mesoscale Alpine Program, autumn 1999, see Bougeault et al., 2001) and ESCOMPTE (Expe´rience sur Site pour Contraindre les Mode`les de Pollution atmosphe´riques et de Transport d'Emissions, summer 2001, see Cros et al., 2004) field experiments, a UHF wind profiler network was deployed in Provence, near the coast, to document the spatial and temporal structure of the flow. The network approach combined with high vertical and time resolutions of the UHF-wind profilers enables the study of the inhomogeneity and unsteadiness of the mistral as well as mistral-atmospheric boundary-layer (ABL) interactions.
In this paper, seven cases of mistral documented by UHF wind profilers are reported: three during MAP, in autumn, and four during ESCOMPTE, in summer. Bordreuil et al. (1973) have shown that the mistral exhibits a seasonal variability on either speed or direction. In cold seasons, the mean mistral is featured by strong northerly wind that can persist over one week and often reach 15-20 m s )1 at coastal ground stations. In warm seasons, the mean mistral is weaker (from 5 to 10 m s )1 ). Its zonal component prevails but rarely persists more than three consecutive days.
The seven mistral events are used to: (i) study the mechanisms responsible for the spatial and temporal variability, sometimes very fast, of the vertical structure, and (ii) investigate the mechanisms that can explain the seasonal variability of mistral characteristics.

Synoptic Environment of the Mistral
The synoptic environment for the mistral is well illustrated by the meteorological situation of the MAP IOP 15 during 06-08 November 1999. It is characterized by the passage of a cut-off low over the North Sea that induces a north-westerly flow impinging on the Alps range ( Figure 2a). Genoa cyclogenesis is then triggered in two phases (Egger, 1972;Buzzi and Tibaldi, 1978;Bleck and Mattocks, 1984). The first phase (Figure 2a) is associated with retardation of the cold front associated with the deep cyclone over the North Sea, and the rapid formation of a shallow vortex supplying a low level source of potential vorticity at the south-western edge of the Alps (Aebischer and Scha¨r, 1998). During the second phase (Figure 2b), the growth rate drops to baroclinic values as a classical cyclogenesis (Bleck and Mattocks, 1984;Tafferner and Egger, 1990). Surface winds strengthen and veer Figure 2. ECMWF analyses from the 06 November 1999 1200 UTC (TU) to the 08 November 1999 00 UTC (given each 12 hours). Surface pressure (hPa) is indicated by sold lines, 500 hPa geopotential height (m) by dashed lines. The coloured scale gives the 1000 hPa horizontal wind speed (WS). Black arrows give the corresponding wind direction. south-eastward. The pressure gradient over the Mediterranean increases with the development of the Azores ridge over Spain and western France (Figure 2c) supplying cold and dry air that passes over the Alps. This results in the displacement of the cyclone towards south-eastern Italy (Figure 2d). Then, the pressure gradient vanishes as the cyclone moves farther eastward. Despite a persistent offshore mistral, the onshore winds stop on the 08 November 1999 1200 UTC. Table I reports the main features of selected MAP and ESCOMPTE mistral cases. Three mistral events occurred during MAP and four during ESCOMPTE. The 1000 hPa ECMWF (European Centre of Medium-range Weather Forecasts) analyses along the AB and CD transects (Figure 1) are used to compute the geostrophic winds from the pressure gradient between the Gulf of Genoa cyclone and the Azores high pressure system. In most mistral events reported in Table I, the configuration of the synoptic environment is the same as described above despite a weaker synoptic forcing. For the weaker case (21-23 June 2001), the cyclogenesis in the southern Alps is a consequence of an interaction between a westerly flow and a overheated land surface with a relative cold sea surface.
In the following, three mistral cases are detailed: the strongest event of the 06-08 November 1999, the moderate event of the 01 July 2001 and the weak event of the 21-23 June 2001.

Experimental Set-Up
The measurements have been made using Degre´ane UHF wind profilers, and consist of the time evolution of the vertical profiles of the three wind components. They are obtained along a single vertical beam and two, or four (depending upon the radar), oblique beams slanted at an off-zenith angle of 17°the half-power beam width being 8.5°. The wind profilers work with a frequency of 1238 MHz (»0.3 m wavelength), and with a peak power of 4 kW. Returned echoes are due to the air refractive index fluctuations advected by the wind. The wind velocity is estimated from the frequency corresponding to the mean Doppler shift obtained in the radar echo. The data quality control and processing are carried out through a consensus algorithm based on time and height continuity of measured spectra. The consensus works over a 60-min period providing a wind profile each 15 min from a height of 0.1-0.3 km AGL up to 2.5-4 km AGL (depths probed by UHF wind profilers are indicative since they are sensitive to meteorological parameters such as humidity). The vertical resolution is typically 75-150 m. The errors in the horizontal (vertical) wind measurements are typically 1-2 m s )1 (0.25-0.5 m s )1 ).  Figure 1 for the exact location of STC and TLN). Both wind profilers were near the shore, TLN being located in an urban area and STC in the countryside. During ESCOMPTE, the STC wind profiler was still available, the MGN radar was located closer to the sea, the MRS radar at Marseille in an urban area and the AIX radar at Aix les Milles in a military airport. On average, those radars were installed 30 km apart. During each experiment, routine radio soundings were launched at Nıˆmes and Lyon by Me´te´o-France ( Figure 1) each six hours in the MAP experiment (each three hours on the 06 November 1999) and each 12 hours in the ESCOMPTE framework. This whole experimental network, i.e. wind profilers and radio soundings, allows a novel mesoscale investigation. On 06 November 1999 from 1200 to 1400 UTC, the mistral starts blowing from the north-west at STC and from the west at TLN between 15 and 25 m s )1 above 1 km AGL. Between 06 November 1999 at 1400 UTC and 07 November 1999 at 0400 UTC at STC, the wind increases up to 25-35 m s )1 and veers to the north/north-west up to the upper range of the UHF measurements. Above TLN, the wind increases on 06 November 1999 between 1400 and 2100 UTC but remains slightly weaker than above STC (i.e. between 25 and 30 m s )1 ) and veers to the west/north-west up to the UHF radar detection limit. The difference in wind direction, which is more eastward at TLN than at STC, is a classical mistral feature.

The Wind Profilers Observations
Between 07 November 1999 at 0400 UTC and 08 November 1999 at 0900 UTC at STC, and between 06 November 1999 at 2100 UTC and 07 November 1999 at 0600 UTC at TLN, strong northerly winds are found below 1 km AGL with a low-level jet structure (LLJ). The wind speed maxima, found around 0.5 km AGL, progressively decrease from 30 to less than 15 m s )1 . One noticeable difference between the two sites is that the mistral event ends seven hours earlier at TLN than at STC.

THE WEAK AND MODERATE SUMMER MISTRAL EPISODES (21-23 JUNE AND 01 JULY 2001)
During ESCOMPTE, two interesting mistral events have been documented by the UHF wind profilers listed in Table II: a weak mistral from 21 to 23 June 2001 and a moderate mistral from 30 June to 1 July 2001. Unfortunately, the UHF wind profiler located at TLN was not operated during ESCOMPTE so the spatial variability of the mistral flow is addressed on a spatial scale of 30 km. Table I indicates that these events are characterised by a weaker pressure gradient than during the MAP IOP 15 event, leading to substantially weaker winds as displayed in  feature is that the mistral is embedded in a 1.5-km deep layer at night between 2200 and 0900 UTC, with maximum wind speed between 0200 and 0600 UTC, whereas during the afternoon, it is 'lifted up' above 1.5 km AGL, as the near-surface wind, i.e. below 1.5 km AGL, decreases down to » 5ms )1 , and veers to the west. This behaviour is systematically observed above the two radar sites during the three days of the event.
For the moderate event (  between 1.5 and 2.5 km AGL above the two sites. During the onset phase and similar to the 21 to 23 June 2001 event, weak westerly winds (i.e between 2 and 10 m s )1 ) blow up to 1 km AGL. Although the time series of the vertical profiles of wind direction profiles are quite similar at AIX and STC, the time series of the vertical profiles of wind speed differ between AIX and STC despite the proximity of the radars (here installed 25 km apart). At STC, as opposed to AIX, the mistral flow is embedded in a 1.5-km deep layer. This suggests that the local structure of the mistral may be combined with small-scale local winds because of the close-by topography and the proximity of the sea. It should also be noticed that the near-surface weak westerly wind is not observed during the afternoon on the 01 July 2001.

Discussion
The purpose of this section consists in interpreting the mistral observations obtained with the wind profilers. The discussion especially focuses on the transition between the deep and shallow mistral during the severe event (06-08 November 1999) and the interactions of the mistral with land-sea breezes during less marked episodes (21-23 June and 01 July 2001).

THE AUTUMN MISTRAL
During the 06-08 November 1999 mistral event, the synoptic forcing is well marked so that the whole Alps ridge is affected by the northerly flows. The linear theory (Smith, 1989) and the RGSW theory (Scha¨r and Smith, 1993) can be used to interpret the UHF wind observations. The incident conditions are defined by the vertical soundings at Lyon (Pettre´, 1982). Figure 6 display the horizontal wind ( Figure 6a) and the isentropes (Figure 6c) interpolated from the radiosonde data at Lyon. The same presentation is made on the radiosonde data at Nıˆmes, located in the southern Rhoˆne valley (Figure 6b and 6d).
The analysis of the flow upstream the Massif Central and the Alps (Figure 6a) clearly shows weaker winds (10-20 m s )1 range below 3 km AGL) and suggests a blocking process. Furthermore, a substantial shift in the orientation of the winds occurs. On 06 November 1999 between 1200 and 1800 UTC, upstream conditions are marked by north-westerly flows that blow below 15 m s )1 . From 06 November 1999 at 1800 UTC to the 07 November 1999 at 0000 UTC, northerly flows prevail. Afterwards northeasterly flows affected the mountain ranges.
The analysis of the Nıˆmes sounding (Figure 6b) shows that the transition between the deep and shallow structure occurs farther westward than STC and confirms that it is not a local feature (see Figure 1). The soundings allow us to better appreciate the vertical structure of the mistral that is limited to 3 km AGL in the wind profiler time series. The setting up of the low-level mistral is associated with a decoupling in the dynamics of the lower and upper troposphere. The wind speed magnitude reported at Nıˆmes is similar than those observed by the two wind profilers (in the 20-30 m s )1 range). Unfortunately, the wind cessation at Nıˆmes is not part of the available observations. The analysis of the isentropes (Figures 6c and 6d) shows that the mistral is associated with a well-mixed layer topped by a very stable layer. The thermal inversion, that corresponds with the maximum vertical isentropic gradient, progressively descends throughout the episode as the mistral depth (Figure 6d).
In the linear theory (Smith, 1989), two parameters control the flow: the horizontal aspect ratio of the obstacle r ¼ a y /a x where a y (or a x ) is the horizontal dimension of the obstacle in a direction perpendicular (or parallel) to the flow. the dimensionless mountain heightĥ ¼ Nh/U where N is the buoyancy frequency, h is the maximum mountain height and U is the incident wind speed.
The values taken by the parameters (r,ĥ) allows the building of the Smith regime diagram for the linear theory. It predicts the occurrence of mountain waves (MW), wave breakings (WB) and flow splitting (FS) in a flow past an obstacle. WB is featured by a stagnation point aloft of the obstacle. When the flow splits around an obstacle, the stagnation point occurs on the windward slope. The linear theory also predicts that a mixture of FS and WB is possible. From the incident wind orientation, different obstacles are encountered by the air masses. Figure 7 displays the upstream topography at STC (Figure 7a) and TLN (Figure 7b) for north-westerly, northerly and north-easterly incident flows. In north-westerly flows, STC and TLN are affected by the Massif Central. For northerly and north-easterly flows, the Alps plays a major role at the two sites. However, the mountain height varies from one site to the other. The values of h are reported in Table III.
The computation of r is based on the map displayed in Figure 8a that indicates the dimensions of the Massif Central and the Alps. The r values are also given in Table III. Figure 8b reports the linear regime of the flow at STC and TLN.
From the 06 November 1999 1200 UTC to 1800 UTC, the flow regime is similar at STC and TLN since both influenced by the Massif Central. The linear theory predicts a MW regime. UHF observations indicate a deep mistral and confirm the similar wind structure above the two sites.
From 06 November at 1800 UTC to 07 November at 0000 UTC, the linear theory predicts MW at STC and the coexistence of WB and FS at TLN. WB in downslope windstorms triggers LLJ windstorms as described by Smith (1985). The earlier setting of the TLN LLJ is thus predicted by the linear theory. From the 07 November 0000 UTC to the 08 November 1999 1200 UTC, the linear theory predicts a FS regime at STC while the conditions are favourable for MW at TLN. The FS encourages the channelling effects within the Rhoˆne valley that triggers the LLJ at STC. Thus, the STC LLJ has not the same origin as the TLN one. At TLN, the MW are difficult to interpret and limit the validity of the linear theory.
After 07 November 1999, three-dimensional processes such as FS play a major role in the flow. Three-dimensional flows are well described by the shallow water theory in case of two-layered flows. Table IV gives the intensity of the thermal inversion computed at Lyon that separates the well-mixed layer of the mistral from a very stable layer aloft (see Figure 6). The thermal inversion strength progressively increases from the beginning of the mistral to the 07 November 1999 0000 UTC and range from 20 to 30 K km )1 for the shallow mistral. Pettre´(1982) uses the RGSW theory with thermal inversion strength greater than 15 K km )1 . The shallow water theory can be thus applied from the 07 November 1999 0000 UTC and completes the description of the linear theory.
In the shallow water theory (Scha¨r and Smith, 1993), the control parameters are: the non dimensional mountain height M ¼ h/H where h is the maximum mountain height and H is the depth of the well-mixed layer of the mistral.  with U the incident wind speed and g¢ the reduced gravity given by g 0 ¼ g hÀh 0 h , where h 0 (or h) is the potential temperature of the upper (or lower) layer on either sides of the thermal inversion.
The shallow water theory predicts the existence of four regimes. Regime I refers to inviscid irrotational flow with no hydraulic jumps. Regime IIa refers to a wake formation without reverse flow due to the hydraulic jump. Regime IIb indicates a wake formation with a reverse flow. Regime III indicates a wake regime due to a pierced fluid surface (with M > 1).  respectively, to inviscid irrotational flow, wake regime without reverse flow associated with hydraulic jump, wake regime with reverse flow associated with hydraulic jump, and to a wake regime due to a pierced fluid surface (from Scha¨r and Smith, 1993).   Table IV reports the values taken by the parameters (M, F) computed from the Lyon radio soundings throughout the mistral event. Figure 9 presents the shallow water flow regime for the 07 November 1999 0000 UTC to the 08 November 1999 1200 UTC. The flow is non dissipative (regime I) on 07 November 1999 0000 UTC above the two sites. At 0600 UTC, the shallow water theory predicts a wake regime with reverse flow (regime IIb) at the two sites. From 07 November 1999 1200 UTC, the theory predicts a wake regime triggered by a pierced fluid surface. As the thermal inversion is located below the mountain crest, the channelling effect upstream STC is more efficient. TLN is in the wake regime predicted by the theory that gives an explanation for the early cessation of the TLN winds (see Figure 3b). The presence of wake is confirmed by Jiang et al. (2003).

Date
This discussion concerns the well-marked mistral situations when the whole Alps ridge is affected by a similar flow. Indeed, the conclusions made up for the MAP IOP 15 mistral event can be transposed to interpret a similar transition deep/shallow mistral that occurred from 16 to 18 June 2001. Figure 9. Shallow water flow regimes at STC and TLN from 07 November 1999 0000 UTC to 08 November 1999 1200 UTC. Each point corresponds with the radiosounding launching at Lyon and is associated with the duration from the onset of the mistral that occurs on the 06 November 1200 UTC to the radio sounding launching.

THE SUMMER MISTRAL
For summer weaker mistral situations of the 21-23 June and 01 July, a linear theory analysis is carried out. However, the time resolution of the upstream conditions at Lyon is coarser (12 h). Figure 10 presents the vertical soundings of the horizontal wind and the potential temperature at Lyon and Nıˆmes for the weak mistral event. For clarity of the figure and since the low-level mistral observed by the wind profilers is stronger during nighttimes, only the soundings at 0000 UTC are displayed. The incident conditions are less pronounced that the autumn event with wind speed less than 10 m s )1 below 2 km AGL (Figure 10a). Hence, the mistral at the exit of the Rhoˆne valley (Figure 10b) is below 15 m s )1 and looks like a LLJ as the wind profilers reveal (Figure 4). Note that the mistral at Nıˆmes is more intense than the mistral reported by the two wind profilers. The analysis of the stratification (Figure 10b) shows that the low-level layers are continuously stratified and no well-mixed layer occurs as in autumn. Thus, only the linear theory is applied to interpret the wind data. Table V gives the linear flow regime for the weak mistral event and Figure 12a reports the results in the linear regime. The theory generally predicts that the flow is governed by the splitting around the Alps and the Massif Central. The nocturnal LLJ observed by the wind profilers ( Figure 5) likely results from the channelling by the Rhoˆne valley. Moreover, the mistral observed at Nıˆmes and by a sodar set up at Avignon during ESCOMPTE located nearer the Rhoˆne valley axis (see Figure 1 for its exact location) is 2m s )1 stronger than the coastal mistral observed at AIX and STC. However, the theory fails to describe the wind structure changes from daytimes to nighttimes.
Firstly, as the synoptic situation is not very pronounced, the Alps is not only affected by incident northerly flow. During daytimes, the coastal mistral is affected by sea breezes and by thermally driven convection over ground, observed by various platforms of the instrumental network. These local thermal effects that are strong enough to lift up the mistral above 2 km AGL (Figure 4) disturb the validity of the linear theory. During nighttime, the land breeze partly accelerates the low-level mistral giving birth to a LLJ. Moreover, nocturnal radiative cooling increases the stratification in the TABLE V Linear theory regime for the STC and AIX sites during the 21-23 June and 1 July mistral events, h is the mountain height, r is the horizontal aspect ratio, Nh/U is the dimensionless mountain height. The regimes are referred to MW mountain waves, WB wave breaking, FS flow splitting (from Smith, 1989 atmospheric boundary layer. It results in nocturnal summer conditions that are more similar in term of stratification with the winter conditions explaining the onset of the nocturnal LLJ. The local thermal effects are not obviously considered in the upstream conditions. Secondly, the incident wind direction profile exhibits low-level wind shears, particularly outlined on 22 June 2001 0000 UTC (Figure 10b) that render difficult the computation of Nh/U and r of the linear theory.
Linear theory can be used to interpret the moderate summer events of the 01 July. Figure 11 displays the vertical soundings of the horizontal wind and the potential temperature at Lyon and Nıˆmes from 30 June 1200 UTC to 02 July 0000 UTC. The mean wind speed measured at Lyon as the mistral is fully developed is around 10 m s )1 below 2 km AGL (Figure 11a). The incident winds veer from north-westerly to north-easterly. The resulting flow observed at Nıˆmes experiences a marked low-level acceleration (Figure 11b) especially at 02 July 0000 UTC when the winds stop at AIX and STC ( Figure 5). The analysis of the potential temperature profiles (Figures 11c and d) shows that the low-level layers are less stable. The well-mixed depth of the mistral is thus clearly visible at Nıˆmes and gets progressively thinner with time (from 2 to 1 km).  Table V gives the linear flow regime for the moderate mistral event and Figure 12b reports the results in the linear regime. The theory predicts a FS regime at the onset of the mistral, consistent with the weak winds below 5ms )1 measured at Lyon. Then, the MW regime corresponds with the deep mistral structure as in the autumn case. FS is predicted at the mistral breakdown. This regime can be linked with channeling effects that trigger a LLJ at STC (Figure 5a) and at Nıˆmes (Figure 11a). However, the theory fails to describe the small-scale effects responsible for the differences in the wind structure of STC and AIX. For instance, the breakdown of the deep mistral at AIX is associated with a LLJ at STC. Strong variability occurs at a 30-km horizontal scale and the weak time resolution of the vertical soundings prohibits the interpretation of the flow with the linear theory. Aircraft data obtained by the Dornier (Corsmeier et al., 2004) show that AIX is located at the limit of the eastern mistral shear line that borders the Alps wake. That feature is also observed during the 28 June 2001 mistral episode . The presence of the wake is predicted by the shallow water theory of Scha¨r and Smith (1993) since the thermal inversion at 2 km AGL pierces the mountain height (2.5 km) in a northeasterly incident flow.

Summary
This paper examines the inhomogeneity and unsteadiness aspects of the mistral LLJ dynamics using a network of UHF wind profilers that was operated during the MAP and ESCOMPTE field experiments that took place in autumn 1999 and summer 2001, respectively. Some interpretations of the observed wind structure are carried out by the use of the linear theory of Smith (1989) and the shallow water theory of Scha¨r and Smith (1993).
Both theories show that the wind structure observed downstream the Alps and the Massif Central results from the time evolution of the upstream conditions especially when synoptic conditions are well marked. They predict that the FS process plays a major role to explain the shallow mistral at STC. The LLJ, triggered by north-easterly upstream winds, is strengthened by the channelling effect explaining the cause of its persistence. The linear theory indicates that the LLJ at TLN results from WB occurring in downslope windstorms. The loss of momentum in associated hydraulic jumps develops a wake with reverse flow explaining the early cessation of the TLN winds. Confirmations of such processes can be found in non-hydrostatic numerical simulations that highlight a tropopause foliation over the Alps during the 06-07 November 1999 mistral event (Hoinka et al., 2003). Thus, the theories evidence that although the STC and TLN low-level jets look similar, their origins are very different.
Simple theories can not be applied for weaker mistral events as local effects play a major role for the wind structure. If the mistral is weak, its LLJ structure is removed by the sea breezes and convection over heated land and lifted up above the atmospheric boundary layer. During nighttime, the summer mistral dynamics are similar to the winter dynamics linked with the advection of relative cold air embedded in the stable boundary layer. Moreover, the nocturnal summer mistral is accelerated by the land breeze. These arguments give possible explanations for its nocturnal LLJ structure.
Our study is a first step to understand the strong time and spatial variability of the mistral. Future work will focus on channelling and downslope processes with the help of the hydraulic theory. Furthermore, high resolution numerical simulations will be performed to describe the interaction between the mistral and the gravity wave activity in the lee of the Alps and the possible role of these waves on the mistral breakdown and also on the mistral/sea-breeze interactions.