1. Introduction
Windstorms associated with extratropical cyclones are among the most destructive and costly natural disasters in midlatitudes. Their dynamics at the synoptic scale have been studied intensively since the development of the Norwegian model (Bjerknes and Solberg 1922). Subsequently, the scientific community has focused on the mesoscale characteristics outlined by the conceptual conveyor belt model (Browning 1971; Carlson 1980). This model describes the mesoscale structure of the cyclone by the evolution of three-dimensional air flows, namely, the cold conveyor belt (CCB), the warm conveyor belt (WCB), the dry air intrusion (DI) and sting jets. According to Hewson and Neu (2015), extreme surface wind gusts are attributed to these different air flows during the cyclone lifetime, and can occur at variable locations and times relative to the cyclone center. The most damaging surface wind gusts are usually due to the sting jet when present, defined as an airflow descending from midlevels to the top of the atmospheric boundary layer (ABL; Browning 2004). Previous studies (Browning 1990; Schultz 2001; Rivière et al. 2020) have shown that the strongest surface winds are associated with the CCB that originates in the lower troposphere on the cold side of the warm front, as it wraps cyclonically around the low-pressure center. In other cases, strong surface winds in mature stage of cyclone development occur ahead of the cold front along the WCB (Martínez-Alvarado et al. 2014; Hewson and Neu 2015).
Gusts represent the wind hazard most likely to be associated with serious and harmful windstorm damage, ranging from human fatalities and injuries to losses of livelihoods (Sheridan 2018; Pinto et al. 2019). They affect a wide range of infrastructures including buildings, power distribution outages, and road and air traffic. As stated by the WMO (2018), gusts are intense, localized peaks of wind, characterized by a short duration, measured over 3 s. These small-scale weather events occur when high-momentum air is brought to the surface. Due to their short duration and localized nature, wind gusts are not comprehensively described by standard meteorological observation networks and are represented by parameterizations in current numerical weather prediction models (Friederichs et al. 2009).
Different processes can be responsible for the transport of strong winds to the surface in windstorms. Using convection-permitting simulations, Ludwig et al. (2015) showed that the strong winds can be due to deep convection along the cold front, which contributed to downward mixing of momentum, thus producing powerful wind gusts at the surface. In the case of the St Jude windstorm involving a sting jet, Browning et al. (2015) demonstrated the importance of convective showers and evaporative cooling in bringing down high momentum from the sting jet to the surface. In a study of an idealized sting-jet cyclone, Rivière et al. (2020) showed that the downward transfer of momentum in the boundary layer is organized by convective rolls.
Boundary layer rolls are one of the most common forms of the convective ABL. Previous studies (e.g., Weckwerth et al. 1999; Young et al. 2002) have widely investigated convective rolls. They are defined as quasi two-dimensional organized vortices, aligned approximately in the wind direction, creating areas of alternating up- and downward motion. Convective rolls play a significant role in transporting momentum, heat, and moisture across the ABL, thanks to their coherent upward and downward branches that extend vertically through the depth of the ABL. Roll formation is attributed to two mechanisms: dynamic instability, which tends to organize convection into rolls, and thermal instability, which forces them to form. These two instabilities are often combined. Modeling and remote sensing studies have shown that rolls are often formed by thermal instability in cold-air-outbreak events (Chen et al. 2019; Brilouet et al. 2017). Doppler radar observations (Morrison et al. 2005; Lorsolo et al. 2008) have revealed the presence of convective rolls spanning the depth of the hurricane boundary layer. These rolls are responsible for transporting high-momentum air from the upper boundary layer to the surface, resulting in higher surface wind speeds. In addition to the transport of momentum, convective rolls can enhance air–sea exchange (Foster 2005; Zhang et al. 2008), leading to anomalous dissipative heating, increased wave breaking and evaporation of sea spray. On the modeling side, high-resolution convective roll simulations are usually idealized studies (Huang et al. 2009; Rivière et al. 2020), hindered by a lack of realistic initial and boundary conditions. This highlights the need to better understand convective rolls in real windstorm situations.
While wind gusts form on small scales, windstorms extend over a few thousand kilometers. The large range of scales involved is a major challenge for windstorm modeling. In most of the cases studies, numerical weather prediction models are used with a limited horizontal resolution of several kilometers at best (Shestakova et al. 2018; Slater et al. 2017). Thus, small-scale processes in the windstorm boundary layer are not resolved and their effects must be parameterized. Few studies have employed realistic large-eddy simulations (LESs)—i.e., simulation with a fine enough grid to explicitly resolve eddies containing most of the energy and responsible for most of the turbulent transport (Pope 2004)—for cyclones and only with nested domains of limited spatial extent, due to limited computational resource (Pantillon et al. 2020). Consequently, these LESs inherit the limitations of the parent model and thus the representation of mesoscale processes is not captured through the rapidly evolving lateral boundary conditions.
The main purpose of this paper is to investigate the small-scale processes responsible for the transport of strong winds to the surface, through the use of a LES in a nested configuration, resolving both the entire windstorm structure and the fine-scale processes, and to discuss their characteristics, transport contributions and sensitivity to the horizontal grid spacing. The windstorm studied is the Mediterranean windstorm Adrian of late October 2018. The Mediterranean Sea has long been considered the most active cyclogenetic area in the Northern Hemisphere in winter, conducive to the generation of very high wind speeds (Petterssen 1956) through orographic channeling and influenced by various factors, including intense air–sea interactions. Unlike midlatitude cyclones that develop over the open oceans, Mediterranean cyclones are characterized by their small size, short duration, and weak intensity (Flaounas et al. 2014). Nevertheless, they often have a severe impact, being responsible for heavy precipitation and extreme winds that affect both the environment and the population of about 500 million people in the Mediterranean basin. The dynamics of these events have been well studied (Flaounas et al. 2022). However, a solid understanding of the processes that transport strong winds to the surface remains crucial for risk prevention and potential economic impacts.
Adrian (also known by the name Vaia assigned by the University of Berlin; Cavaleri et al. 2019; Davolio et al. 2020; Giovannini et al. 2021) formed west of Sardinia in the western Mediterranean Sea on 29 October 2018, as a result of a cold inflow from the Gulf of Lion. A strong contrast between the cold polar air and the warm, moist marine boundary layer led to the rapid intensification of the cyclone, which underwent explosive cyclogenesis before moving rapidly northward. With wind gusts over 180 km h−1 in Italy, and in addition to the serious loss of life, the impact on the environment was dramatic: 41 000 ha of forest were damaged and thousands of trees were uprooted (Forzieri et al. 2020). On Corsica, southwest wind gusts of more than 140 km h−1 and at 176 km h−1 were recorded at the coastal stations of Ajaccio-Parata and Marignana, respectively. Although the human toll was limited to one injured person in France, the damage to port and coastal infrastructures was significant, as they were severely affected by the mechanical shocks induced by the action of destructive winds and waves.
Below, the modeling strategy and analysis tools are described in section 2. Section 3 provides a general overview of the mesoscale dynamics of the case study as well as the validation of simulations against in situ measurements, focusing on the wind. The characteristics of the boundary layer winds and the fine-scale processes responsible for momentum transport are discussed in section 4. Subsequently, section 5 is devoted to the sensitivity analysis of fine-scale processes to horizontal resolution. Section 6 summarizes and discusses the main results.
2. Methods
a. Numerical simulations
Simulations of windstorm Adrian are performed with the Meso-NH nonhydrostatic atmospheric model (Lac et al. 2018). Their vertical grid has 72 levels up to 22 km, with a finer spacing of 10 m at the first level. To simulate the mesoscale conditions of Adrian during its life cycle, as well as the origin of strong winds at this scale, we run a simulation (Meso1000 thereafter) with a 1000-m horizontal grid spacing allowing an explicit representation of the deep convection, centered on the western Mediterranean, i.e., a region of 750 km × 750 km (D1 domain in Fig. 1). Meso1000 is initialized at 0600 UTC 29 October 2018, from the European Centre for Medium-Range Weather Forecasts (ECMWF) operational analysis and run for 15 h using boundary conditions from ECMWF analysis at 1200 and 1800 UTC and 3-h forecasts at 0900, 1500, and 2100 UTC. Meso1000 uses the fifth-order weighted essentially nonoscillatory (WENO) advection scheme (Shu and Osher 1988) for momentum variables and the piecewise parabolic method (PPM) advection scheme (Colella and Woodward 1984) for scalar variables. Meso1000 uses a 1.5-order closure scheme for turbulence (Cuxart et al. 2000) set in 1D mode. In this case, the horizontal gradients of momentum, heat, moisture, turbulent kinetic energy (TKE) and their turbulent fluxes are assumed to be negligible compared to their vertical counterparts. Only the latter are thus taken into account to estimate the turbulent fluxes, which is a reasonable assumption for a convection permitting resolution, such as in Meso1000. The mixing length is parameterized as the maximum vertical displacement allowed by an air parcel characterized by a TKE value (Bougeault and Lacarrère 1989, BL89 thereafter). Meso1000 includes an eddy-diffusivity mass flux (EDMF) parameterization for dry thermals and shallow convective clouds (Pergaud et al. 2009), a single-moment microphysical scheme for mixed-phase clouds (Pinty and Jabouille 1998), the two-stream scheme of Fouquart and Bonnel (1986) for shortwave radiation, the Rapid Radiative Transfer Model (Mlawer et al. 1997) for long wave radiation and the Surface Externalised (SURFEX) scheme for the exchange of surfaces with the atmosphere (Masson et al. 2013) using the Coupled Ocean–Atmosphere Response Experiment (COARE) 3.0 bulk algorithm (Fairall et al. 2003) to represent the turbulent fluxes at the air–sea interface. Topographic data have a horizontal resolution of 250 m and are smoothed to avoid numerical instabilities on the steepest slopes of the Corsican mountains.
Domain D1 of the Meso1000 simulation, with terrain height color shaded. Domain D2 of the Les200 simulation is nested in D1 and centered on the northwestern Mediterranean Sea. D3 refers to the domains of both the Les100 and Les50 simulations nested within D2.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
To study the fine-scale processes responsible for the transport of strong winds to the surface, and to resolve the different scales involved in a storm, a reference LES (Les200 thereafter) nested in Meso1000 is performed over the same period with a horizontal grid spacing of 200 m over a region of 400 km × 400 km covering a part of Mediterranean Sea and the whole of Corsica (D2 domain in Fig. 1). It was possible to run this LES over a domain wide enough to capture both the mesoscale dynamics and the fine-scale processes thanks to the parallel computing capability of the Meso-NH model (Pantillon et al. 2011). To verify the ability of the Les200 simulation to resolve relevant fine-scale processes, additional nested simulations are used to achieve even higher horizontal resolution. Two simulations centered on the region of strong winds over the sea, i.e., a region of 100 km × 100 km (D3 domain in Fig. 1), with horizontal grid spacings of 100 and 50 m (Les100 and Les50) are integrated over 30 min from 1500 to 1530 UTC. Note that one-way grid nesting between D1-D2 and D2-D3 is applied, i.e., lateral boundary layer conditions are provided by the outer model for the inner model at every time step. Les200, Les100, and Les50 use the same schemes as Meso1000 with the exception of three. First, the EMDF scheme is turned off. Second, a fourth-order centered advection scheme is used for momentum, which is more accurate, less diffusive, and provides better effective resolution. Third, the turbulence scheme is used in its 3D version and with the Deardorff (1980) mixing length, which represents the characteristic size of the most energetic eddies equivalent to the size of the grid. A summary of the settings is shown in Table 1.
Configuration of the Meso-NH experiments.
To evaluate the simulated cloud fields, reflectances are computed from the model outputs and directly compared to the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) observations on board Meteosat Second Generation (MSG). In the following, the High Resolution Visible (HRV) channel that covers the 0.4–1.1-μm spectrum is adopted to benefit from its higher spatial resolution (1 km at nadir) compared to other channels (3 km at nadir). The synthetic reflectances are computed using the interface included in Meso-NH (Chaboureau et al. 2008) with the Radiative Transfer for the Television and Infrared Observation Satellite (TIROS) Operational Vertical Sounder (RTTOV) code, version 13.2 (Saunders et al. 2018).
To investigate the airflow causing the high winds in Adrian, Lagrangian trajectories introduced by Gheusi and Stein (2002) are used in the Meso1000 and Les200 simulations. The Lagrangian trajectories are calculated online by initializing passive tracers with their initial 3D coordinates at each grid point in the simulation domain. The tracers are advected by the PPM scheme, which is known to conserve the mass properties of the tracers with low numerical diffusion. Consequently, the initial position of every air parcel is known throughout the domain at each time step, i.e., every 10 s. Back-trajectories are then calculated offline for a chosen period of interest. This method then makes it possible to know the origin of the air masses and to follow their positions over time.
b. Autocorrelation function
3. Strong wind at the mesoscale
The purpose of this section is to relate the mesoscale structure of the strong winds to the air flows. The wind speeds at 10 m over the D2 domain for the Meso1000 and Les200 simulations are shown at 1515 UTC when the cyclone approaches Corsica (Figs. 2a,b). For both simulations, wind speeds up to 40 m s−1 occur below the occluded front, and are confined to a narrow belt on the southern flank of cyclone, over the sea. Figure 2c shows the thermal structure (shadings) of the low pressure system (gray contours). Trajectories are calculated between 0915 and 1515 UTC and are then selected based on a wind speed threshold above 40 m s−1 below 800-m altitude at 1515 UTC. A coherent bundle of trajectories shows a qualitatively similar behavior in their evolution along the flow. Beginning from the Gulf of Lion near the cold side of the warm front, the trajectories follow the cold side of the curved front. They wrap cyclonically around the southern flank of the low pressure center at 1515 UTC. During the considered period (Fig. 2d), the wind speed along trajectories ranges between 10 and 20 m s−1 around 0915 UTC, until reaching speeds of about 48 m s−1 around 1515 UTC in the elongated region of strong winds while their height remains in the lower troposphere as a low-level jet. This is a typical behavior of the CCB (e.g., Clark et al. 2005; Schultz 2001) that is the origin of the strong winds in windstorm Adrian. Note that trajectories do not indicate the presence of a sting jet that would originate in the midtroposphere.
(a),(b) Horizontal wind speed at 10-m height (in m s−1) at 1515 UTC from (a) Meso1000 and (b) Les200 simulations over the D2 domain. The black and white squares represent the zoomed area shown in Figs. 6 and 10, respectively. (c) Potential temperature at 925 hPa (in K) at 1515 UTC from the Meso1000 simulation over the D1 domain. A sample of back trajectories reaching horizontal wind speed values above 40 m s−1 below 800 m at 1515 UTC is illustrated with blue lines. MSLP is shown with gray contours every 2 hPa in (a)–(c). (d) Evolution of height and wind speed along Lagrangian trajectories from 0915 to 1515 UTC. The median and the interquartile range are shown with the bold line and shading, respectively, for all ≈40 000 trajectories reaching the same threshold as in (c).
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
Cloud structures associated with Adrian development are studied using normalized reflectances of the SEVIRI HRV channel. Observed and simulated reflectances are shown at 1515 UTC when Adrian turned to the northeast, while approaching Corsica (Fig. 3). In the observation, large and thick stratiform clouds cover the CCB region in the northwest of the domain (Fig. 3a). To the west of Corsica, high clouds roll up cyclonically in the form of a comma indicating the center of the low pressure system. The position of Adrian (magenta dotted line) is estimated every 15 min by following the center of the cloud roll-up. The track indicates that the cyclone formed in the morning over the sea to the west of Sardinia. The roll-up is accompanied by small isolated clouds shaped like cumulus (in white) indicating deep convection. We find the same type of structures in the south, close to Sardinia. In the simulations, the stratified cloud cover is correctly reproduced, although the cloud roll-up is less pronounced regardless of the horizontal resolution (Figs. 3b,c). This is partly due to the missing high reflectance values attributed to unsimulated 3D solar effects in the RTTOV code, e.g., shadow or reflection effects. Unlike the isolated clouds to the west and north of Corsica, those in the south of the domain are well reproduced by both simulations. Thanks to the fine resolution, Les200 shows smaller and well-defined cloud structures. The position of Adrian is estimated every hour in both simulations from the minimum pressure over the sea. At 1500 UTC, close to Corsica, the minimum reaches its lowest value at 976 hPa. The simulated roll-up coincides with the position of the minimum sea level pressure. The tracks are relatively comparable and close to the one estimated from observations, despite a slight position error of about 50 km westward at 1500 UTC. In the following, most of the figures will be presented at 1515 UTC.
Normalized reflectance of the SEVIRI HRV channel over the D2 domain at 1515 UTC from (a) MSG observation and (b) Meso1000 and (c) Les200 simulations. The magenta dotted line shows the cyclone track. In (a), the position of Adrian is estimated every 15 min as the center of the cloud roll. In (b) and (c), it is defined every 1 h as that of the MSLP minimum, whose value in hPa is shown in red. In (b) and (c), the cyan contours show the MSLP every 1 hPa below 980 hPa.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
To evaluate the change in wind between 0600 and 2100 UTC over both sea and land, the evolution of the 99th percentile of wind speed at 10 m is calculated over the D2 domain for Meso1000 and Les200 (Fig. 4). Over sea (dashed curves), the 99th-percentile wind speeds increase progressively from 0600 UTC onward. They reach 25 m s−1 at 1000 UTC due to the formation of the low pressure system in the morning. Between 1000 and 1800 UTC, wind speeds exceed 25 m s−1. The wind peak is around 1500 UTC, with maxima above 30 m s−1. As Adrian continues its northward trajectory, the wind speed gradually decreases until the end of the day. Over land (solid curves), an increase in wind speed is noted from 1200 UTC onward, due to the passage of the low pressure system near Sardinia. The wind peak is found between 1600 and 1700 UTC, which indicates the arrival of the windstorm over Corsica. Although the mesoscale structure remains comparable between the two simulations over sea, Meso1000 unexpectedly shows a more prominent peak with a positive difference of about 1.5 m s−1 compared to Les200. An explanation for the more prominent peak is provided in section 5b, dealing with the fine-scale processes that differ between Meso1000 and Les200. On land, the contrast is more pronounced with a negative difference peak of about −5 m s−1 in Meso1000. As shown in Figs. 2a and 2b, winds over Corsica follow the terrain, and are stronger in Les200 than in Meso1000. This is explained by the finer representation of the terrain in Les200.
Temporal evolution of the 99th percentile of wind speed at 10 m for Meso1000 (red) and Les200 (green) simulations over the land (solid) and sea (dotted) parts of the D2 domain.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
To assess the impact of Adrian on Corsica and to evaluate the capacity of the model to reproduce the observations, we use the wind speed at 10 m averaged every 6 min recorded by the Météo-France surface weather network from 0600 to 2100 UTC. For a fair comparison, instantaneous model winds are also averaged over 6 min. Figures 5a and 5b confirm that the passage of Adrian strongly impacted the coastal areas of western Corsica. On land, both simulations are consistent with the wind measurements recorded in the 27 stations. Wind speed is better captured by Les200 than Meso1000 with a correlation coefficient of 0.65 and 0.42, respectively, between simulated and observed values. When considering the 14 exposed stations with records above 20 m s−1, both simulations underestimate the observed maximum wind speed but Les200 also performs better with a negative bias of −2.2 m s−1 compared to −5.2 m s−1 for Meso1000. Over the sea, the footprint of the wind is marked by speeds of up to 40 m s−1. The very high resolution of Les200 makes it possible to visualize striped structures. Note that the northeastern stations also recorded high wind speeds especially in the morning. Contrary to the winds in the western region, these strong winds are caused by the WCB ahead of the cold front to the east of Corsica (not shown). Convective clouds embedded in the WCB led to heavy precipitation in the northeast of Corsica in the morning. In this study, we focus on the strong winds that impacted western Corsica, and that are mainly caused by the CCB, rather than those generated by the WCB that also affected Italy.
(a),(b) Maximum wind speed at 10 m observed on Corsica stations (filled dots) and simulated from (a) Meso1000 and (b) Les200. The white square represents the zoomed area shown in Fig. 10. (c),(d) Temporal evolution of wind speed at 10 m at the (c) Ajaccio-Parata and (d) Marignana stations measured every 6 min, and averaged every 6 min from Meso1000 (red) and Les200 (green) simulations at the grid points closest to the stations. The results are valid between 0600 and 2100 UTC 29 Oct 2018.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
Time series at the Ajaccio-Parata and Marignana stations (Figs. 5c,d) further illustrate the arrival of Adrian on western Corsica. Observations and simulations show that 10-m wind speeds were low in the morning, except for the passage of the cold front that was preceded by strong winds recorded around 1000 UTC corresponding to the WCB jet. This sudden increase coincided with a decrease in pressure at that time (not shown). The development and approach of the cyclone results in high speeds from around 1200 UTC onward. The arrival of the windstorm is marked by wind speeds up to 30 m s−1 recorded between 1500 and 1700 UTC in Ajaccio-Parata and 1600 and 1700 UTC in Marignana. After the peak, the wind speed gradually decreases until the end of the day. The time series from Meso1000 to Les200 present differences but a similar overall evolution compared to the wind records. Although there is some variability, both simulations also present a similar pressure evolution with a well-marked minimum (not shown). In Ajaccio-Parata (Fig. 5c), the simulated winds are very close to each other except for the peak. Meso1000 matches the recorded peak value better than Les200 with a negative bias of −1.5 and −2.2 m s−1, respectively. This station located close to the sea shore is strongly influenced by the sea for which the simulated wind at 10 m is stronger in Meso1000 than in Les200 as explained in section 5b. In Marignana (Fig. 5d), Meso1000 underestimates the wind speed during most of the day with a large bias of −9 m s−1 on the peak value, while Les200 well captures the intensity with a small bias of −0.5 m s−1 on the maximum wind speed. This station is a good archetype of the exposed stations described in the statistics above, which explains the higher Les200 score. Compared to the observations, both simulations underestimate wind speeds. These underestimations are much greater in Meso1000, and may be due to a poor representation of subgrid scale processes and orography.
4. Strong wind at the fine scale
a. Presence of strong wind structures
The strong wind region on the southern flank of the low pressure center at 1515 UTC (see black square in Figs. 2a,b) is investigated in Meso1000 and Les200 (Fig. 6) by inspecting horizontal cross-sections at 200-m height (left panels) and vertical cross sections of wind speed (right panels). The horizontal cross sections are over a 20 km × 20 km domain. The vertical cross sections are about 15 km long with northwest–southeast orientation, approximately perpendicular to the mean wind direction. The sections are chosen at the time when the wind speed is maximum over the sea, as shown in Fig. 4. In the horizontal section (Fig. 6a), a smooth and homogeneous zone of wind higher than 40 m s−1 is found in Meso1000. In the vertical section (Fig. 6b), the atmosphere is cloud free and the thick red line indicates a boundary layer height between 500 and 600 m, covered by mesoscale subsidence. The high wind speeds are mainly confined to the boundary layer, with a maximum of 48 m s−1 around 400 m, near the top of the boundary layer. These high wind speeds are a result of the CCB originating in the lower altitudes of the troposphere, as demonstrated in Fig. 2c.
(left) Horizontal and (right) vertical cross section of horizontal wind speed at 1515 UTC from (a),(b) Meso1000 and (c),(d) Les200. In (a) and (c), the horizontal cross section is located at 200-m height and the black line shows the location of the vertical cross section. The blue box represents the area where Les200 diagnostics are calculated. The wind direction is represented by the red arrows. In (b) and (d), the thick red line shows the height of the boundary layer. The black contours show the vertical velocity, where the solid lines correspond to 1 m s−1 and the dotted lines correspond to −1 m s−1. In (d), the magenta contours indicate the mixing ratio of the clouds at 0.1 g kg−1.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
In Les200, the horizontal section (Fig. 6c) reveals the presence of band-like wind structures with velocities greater than 48 m s−1, approximately aligned with the wind direction (indicated by the red arrows) oriented from the southwest. The following figures are all shown at z = 200 m, because the structures are well marked at this altitude. As can be seen in the vertical section (Fig. 6d), the vertical extent of these structures coincides more or less with the height of the boundary layer, ranging from 600 to 1000 m. The difference in boundary layer height between Meso1000 and Les200 can be explained by differences in both mesoscale and fine-scale circulations. The horizontal spacing between two consecutive wind structures is approximately 2 km. The vertical velocity field (black contours) is visualized as a regular alternation of updrafts and downdrafts, which refers to a rotational motion forming boundary layer rolls. Note that such structures are also found at other positions and times in the afternoon along the cyclone trajectory in strong wind areas. In the high wind region, the vertical velocity is as strong as ±3.5 m s−1. Moreover, the region is characterized by an important wind shear that leads to a dynamic instability. At 10 m above the sea surface, the air temperature varies between 10° and 15°C, while the sea surface temperature reaches ≈22°C, resulting in an air–sea temperature difference of about 10°C. This difference indicates that an intense air–sea interaction is occurring as the dry, cold CCB air from the Gulf of Lion is heated and humidified over the warm sea surface. The interaction induces an important heat transfer resulting in sensible heat fluxes reaching 400 W m−2 locally (not shown). In addition, stripe clouds following roll wind structures are present above the boundary layer capping the updrafts. Convective rolls are often described in the literature in relation to parallel lines of clouds, which are created by condensation of moisture transported aloft by updrafts from the roll (Etling and Brown 1993).
Trajectories are computed between 1500 and 1515 UTC and selected from the 2% highest and lowest values of vertical velocity for updrafts and downdrafts, respectively, in the black square shown in Figs. 2a and 2b. The temporal evolution shown in Fig. 7a reveals a symmetry between the updrafts (red curves) and the downdrafts (blue curves), which relatively rapid vertical motion starts after 1510 UTC and lasts only a few minutes. The downdrafts originate from altitudes around 700 m, while the updrafts start from about 200 m of altitude (Fig. 7b). The temporal evolution of wind speed along the Lagrangian trajectories (Fig. 7c) shows that the wind is stronger and remains approximately constant during downdrafts, thus the descending branch transports air of higher momentum from the conveyor belt downward. In contrast, the wind is weaker and slows down during updrafts, thus the ascending branch transports near-surface air with lower momentum upward. This is consistent with the vertical cross section in Fig. 6d, where the regions of strong winds correspond to descent and weak winds correspond to ascent. The potential temperature of both updrafts and downdrafts (Fig. 7d) increases progressively, likely due to the intense heat fluxes over the warm sea. The warming is stronger at low levels of the ABL, where the air becomes warmer than the air above, thus lighter, and undergoes an updraft at around 1510 UTC, contrary to the air coming from the high levels of the ABL, which undergoes a downdraft.
Evolution of (a) vertical velocity, (b) height, (c) wind speed, and (d) potential temperature along Lagrangian trajectories from 1500 to 1515 UTC in Les200. Trajectories are selected from the 2% highest (updrafts) and lowest (downdrafts) values of vertical velocity at z < 1 km at 1515 UTC in the blue box shown in Fig. 6c. The median (thick lines) and the interquartile ranges (shading) are shown for the ≈1000 trajectories selected as updrafts (red) and downdrafts (blue).
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
The different elements discussed above reveal that the thermodynamic conditions are favorable for the development of quasi-two-dimensional convective rolls in the Adrian case.
b. Characteristics of strong wind structures
The spatial characteristics of the organized structures assumed as convective rolls is investigated using the ACF analysis described in section 2b. The two-dimensional autocorrelation of the vertical wind speed field is calculated at 1515 UTC over the domain shown in Fig. 6c. At 200-m altitude, the set of points corresponding to the integral length scale Les reveals an ellipsoidal geometry (brown ellipse in Fig. 8a), characterized by a flatness parameter of 0.83. The direction of the structure (brown dotted line) is approximately parallel to the mean wind direction (blue dotted line), with a slight offset of 6° clockwise. In the direction perpendicular to the roll, the ACF indicates a length scale Los of 2400 m (Fig. 8b). This corresponds to the size estimated visually from the vertical sections in Fig. 6. The aspect ratio, i.e., the ratio between the length scale of the roll and the depth of the ABL (approximately 800 m on average in the domain of Fig. 6c), reaches 3 here.
(a) Two-dimensional autocorrelation over the high wind region shown in Fig. 6 at 1515 UTC. The brown and the blue dashed line indicate the direction of convective rolls and the wind direction, respectively. (b) 1D autocorrelation calculated in the perpendicular direction to the rolls, as a function of spatial distance from the center. Los indicates the characteristic size (length scale) of the roll, considering both ascending and descending branches. Los is calculated as the distance between two autocorrelation maxima.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
From the vertical profiles of the flatness parameter f and the wind and structures direction (Figs. 9a,b), we distinguish three classes of structures and set empirical thresholds on f chosen according to the ACF results and the angle between wind and structure direction. Over a vertical extent of 1200 m, the average wind direction varies clockwise from 57° to 47°, while the direction of the structures varies counterclockwise from 47° to 77°. Elliptical structures are well marked with f > 0.8 (blue stars) below 350 m and mainly oriented in the wind direction with a slight difference of less than 10°. Above the boundary layer, disorganized structures (i.e., nonelliptical) are characterized by f < 0.6 (green stars) and an angle of more than 20° with respect to the wind direction. Structures with f between 0.6 and 0.8 (orange stars) mark the transition between rolls and disorganized structures, and are characterized by a direction difference between 0° and 20°. This is consistent with the observational findings of convective rolls in the hurricane boundary layer by Foster (2005), who found that the roll orientation ranged from 10° to 20° counterclockwise with respect to the mean wind direction. Similarly, in a shallow convective situation, the roll orientation is typically 15°–20° counterclockwise (Atkinson and Wu Zhang 1996).
(a) Vertical profile of the flatness parameter f. Structures organized as rolls (f > 0.8) in blue stars, disorganized (i.e., nonelliptical) (f < 0.6) in green stars, and transition regime (0.6 < f < 0.8) in orange stars. (b) Vertical profile of the mean wind direction and the direction of structures identified by the autocorrelation function. The dashed red line and the gray lines indicate the boundary layer height averaged over the selected domain and the standard deviation of the height, respectively.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
The characteristics inferred from the ACF analysis prove that the organized boundary layer structures are indeed convective rolls. This organization is well marked below z = 350 m, i.e., below the maximum wind of the CCB. Thanks to their coherent up- and downdrafts, the convective rolls facilitate the mixing of the momentum in the ABL and greatly contribute to the downward transport of strong winds by the descending branches.
c. Fate of convective rolls over land
During its crossing of the northwestern Mediterranean Sea, windstorm Adrian approached the west coast of Corsica at 1600 UTC, accompanied by high wind speeds as shown in Fig. 4. One may wonder if the convective rolls found throughout the evolution of Adrian over sea persist over land. Following the same approach as at sea, Fig. 10a shows a horizontal cross section of wind speed at a height of 200 m above ground level over a 20 km × 20 km domain covering a small area on the west coast of Corsica (see white square on Figs. 5a,b), and a vertical section across the coast (Fig. 10b). The horizontal section highlights a distinction in wind structure organization between land and sea. In the coastal region, the organized wind roll structures break up and turn into disorganized structures. The highest wind speeds in the region are found on the coastal hills and mountains. On the vertical section, a similar organization of the structures as in Fig. 6d is found over sea, with a well-marked alternation of updrafts and downdrafts. On land, the winds lose their roll shape and the alternation between positive and negative vertical wind speeds is no longer apparent. Furthermore, the winds undergo orographic uplift and down-lift. The ACF analysis indicates that there is no spatial periodicity in the vertical velocity field over land. Finally, an excess of TKE is produced over land compared to low TKE values at sea, as well as high TKE values over the orography (Figs. 10c,d).
(a),(c) Horizontal and (b),(d) vertical cross sections of horizontal wind speed and turbulent kinetic energy at 1600 UTC from the Les200 simulation. In (a) and (c), the horizontal cross section is located at a height of 200 m above the ground. The black contours represent the topography at 0 (coast of Corsica), 100, 300, and 500 m of altitude. In (b), the black contours shows the vertical velocity, where the solid lines correspond to 1 m s−1, the dotted lines correspond to −1 m s−1, the magenta contours shows the mixing ratio of the clouds at 0.1 g kg−1.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
These results show that the rolls persist until coastal regions and then break up over land. Since open water is a relatively smooth and homogeneous surface, the formation of convective rolls is facilitated. In contrast, several mountain ranges border the sea and increase the surface roughness and its spatial variability. Additionally, as opposed to the strong sensible heat fluxes over the warm sea, the values of these fluxes are negative over land due to the cold surface. Altogether, the irregular topography of Corsica and the altered thermodynamic conditions of the ABL control the evolution of rolls and prevent their occurrence. Strong winds along the west coast of Corsica are susceptible to being transported by a combination of factors. The presence of steep mountain ranges suggests the involvement of orographic forcing in this transport. Furthermore, turbulence could also contribute to this transport, due to the roughness of the region.
5. Sensitivity of fine-scale processes to horizontal resolution
a. Representation in large-eddy simulations
In section 4, it is shown that roll structures are resolved in Les200 and not in Meso1000. Here, the effect of horizontal grid spacing is examined to ensure that the characteristics of resolved structures in Les200 are realistic and not solely dependent on horizontal resolution. Figure 11 displays the same horizontal and vertical sections as Fig. 6 for the Les100 and Les50 simulations. For a fair comparison, the fields are block averaged at the resolution of Les200. The two horizontal sections reveal structures oriented in the wind direction, resembling those shown in Fig. 6c, although a little less pronounced because of the additional small eddies resolved at higher resolutions. As in Les200, the height of the ABL varies between 600 and 1000 m in Les100 and Les50. Apart from some minor details, the influence of horizontal resolution is hardly noticeable. The instability parameter ζ = −zi/L, (where zi is the ABL height and L is the Monin–Obukhov length) is widely used to predict roll convection. The ζ values are 0.5, 0.52, and 0.54 for Les200, Les100, and Les50, respectively. This is consistent with the observational study of Weckwerth et al. (1997) which indicates that organized roll structures appear for ζ < 10 and can persist up to ζ ≈ 25. In the vertical sections, strong winds are also limited to the ABL and organized into coherent structures, while the alternating motion of updrafts and downdrafts as well as the cloud stripes reproduced. Only the shape of structures is less obvious and the horizontal velocity field less organized in Les100, whereas Les50 is more similar to Les200. Horizontal and vertical wind speeds show moderate anticorrelation in all three simulations. Anticorrelation is overestimated in Les200 with a correlation coefficient of −0.45 compared to −0.35 for Les100 and Les50, which may lead to an overestimation of momentum transport. Applying ACF to averaged fields also reveals structures organized in rolls. These rolls have a characteristic size of around 2000 m, while Les200 has a slightly larger size of 2400 m. The flatness parameter is 0.8 for Les50 and 0.83 for Les100 and Les200, and is therefore weakly affected by the horizontal resolution. It is concluded that the organization of the ABL into convective rolls is found in all simulations with low sensitivity to horizontal resolution, thus it is adequately represented in Les200.
As in Fig. 6, but for (a),(b) Les100 and (c),(d) Les50. The fields are block averaged at 200-m resolution.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
Another way to investigate the effect of horizontal grid spacing is to assess whether the most energetic eddies are explicitly resolved, i.e., whether the resolution is sufficient to represent the small-scale dynamics. Smaller and smaller eddies are resolved as the horizontal resolution increases (Honnert et al. 2011). According to Pope (2004), it is admitted that a simulation is qualified as LES if at least 70%–80% of the TKE is explicitly resolved. The quality of the different simulations described in section 2a is assessed in terms of the ratio of resolved TKE to total TKE. As illustrated in Fig. 12a, whatever the resolution considered, the turbulence is mostly parameterized in the first tens of meters of altitude. This is because the eddies near the surface are smaller than the grid spacing. At 200 m of altitude, where the convective rolls are well marked, 70%, 80% and more than 80% of the TKE is resolved in Les200, Les100, and Les50, respectively. The Les200 simulation is then qualified as LES above 200-m altitude.
(a) Ratio of resolved TKE to total TKE for the Les200, Les100, and Les50 simulations. The profiles are averaged over the blue box shown in Fig. 6c. (b) Kinetic energy spectra calculated for the horizontal wind at z = 200 m at 1515 UTC, gray lines correspond to −2/3 theoretical slope of the inertial subrange. The black vertical line corresponds to energy peaks at a wavelength equal to 1000 m. The green, pink, purple, and blue lines correspond to Meso1000, Les200, Les100, and Les50, respectively.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
For the same purpose, the distribution of kinetic energy between different scales is studied using a spectral analysis of the horizontal components of the wind at 200-m altitude (Fig. 12b). Energy spectra are used to determine whether a scale is correctly represented and to define the effective resolution, which is the smallest scale well represented in the model. Energy peaks corresponding to the dominant wavelengths are observed at around 1000 m in the Les100 and Les50 simulations. In Les200, the peak is slightly shifted by 50 m toward the large scales. These peaks are consistent with the characteristic size of a single roll branch (up- or downdraft) derived from ACFs. For Les50 and Les100, the spectrum mainly follows the theoretical slope of the inertial subrange for scales below 1000 m. The effective resolution, which is given by the scale at which the model deviates from the theoretical slope (Skamarock 2004) is around 200 and 400 m in Les50 and Les100, respectively. The spectrum in Les200 marginally follows the theoretical slope of the inertial subrange for wavelengths < 1000 m, with an effective resolution of around 800 m. In contrast, the Meso1000 spectrum drops rapidly. Structures with a size of less than 10 km are not resolved, which is consistent with the absence of rolls. For the three LESs and Meso1000, the effective resolution is about 4 and 6 times the horizontal grid spacing, respectively, due to the different advection scheme and in agreement with previous studies (Ricard et al. 2013).
The results show that all the three LESs are close visually as well as statistically. For this case study, a grid spacing of 200 m is sufficient to represent the convective rolls of about 1000 m in size responsible for the strong wind transport in windstorm Adrian. This conclusion agrees with Thurston et al. (2016), who showed that a horizontal grid spacing of less than 600 m is required to model boundary layer rolls observed under severe fire weather conditions.
b. Downward transport of strong wind
Figure 13 shows the vertical profiles of wind speed and vertical fluxes of momentum in the mean wind direction at 1515 UTC, averaged over the blue square indicated in Fig. 6c. Above z = 400 m, wind profiles differ significantly between Meso1000 and LES. Meso1000 has a weaker wind peaking near z = 400 m. LESs show a gradual increase in wind from z = 1200 m (with stronger values found in Les200), reaching a maximum between 400 and 800 m, which corresponds to the altitude of the CCB. Below z = 400 m, all four simulations show a similar decrease of wind. Vertical momentum fluxes (Fig. 13b) show a progressive strengthening from the top of the boundary layer downward. Near the surface, the vertical transport is comparable between LESs, while it is much stronger and seems unrealistic in Meso1000. This overestimation justifies the larger peak in the temporal evolution over the sea in Meso1000 (Fig. 4). It suggests limitations in the subgrid-scale parameterization schemes used in Meso1000.
Results at 1515 UTC for the Meso1000, Les200, Les100, and Les50 simulations. Vertical profile of (a) horizontal wind speed and (b) vertical fluxes of momentum in the mean wind direction. The profiles are averaged over the blue square shown in Fig. 6c.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
These results are complemented with probability density functions of wind speed values in the different simulations (Fig. 14). The wind speed distributions in Meso1000 are narrower compared to the LES distributions, which are wider and show higher extreme values. At the 400- and 200-m altitudes, the mode in Meso1000 is weaker compared to the modes in the LESs, while it is stronger at z = 10 m, i.e., the wind speeds are higher at the surface. This is consistent with the presence of a more prominent peak for Meso1000 compared to Les200 in Fig. 4 as well as the stronger vertical transport in Meso1000 in Fig. 13. For all three LESs, although the distributions can be considered similar in terms of peak position and skewness, Les200 tends to produce stronger winds at z = 400 m and z = 10 m.
Results at 1515 UTC for the Meso1000, Les200, Les100, and Les50 simulations. Probability density functions of the horizontal wind component at heights (a) 400, (b) 200, and (c) 10 m.
Citation: Monthly Weather Review 151, 10; 10.1175/MWR-D-23-0099.1
6. Conclusions
On 29 October 2018, a Mediterranean extratropical cyclone named Adrian formed to the west of Sardinia, due to a strong contrast between cold polar air and the warm sea surface. Adrian produced strong winds with gusts up to 180 km h−1. As the cyclone crossed the northwestern Mediterranean Sea, the violent gusts caused significant material damage and severe economic losses in Corsica. As wind gusts occur when high-momentum air is transported to the surface, this paper aims to investigate the fine-scale processes leading to the downward transport of strong wind and the formation of maximum surface wind gusts using a large-eddy simulation (LES). The characteristics and contributions of these processes to the wind transport are discussed, as well as their sensitivity to horizontal grid spacing. Previous numerical studies have simulated such processes at relatively coarse resolution for real cases and LES for idealized cases.
The formation of strong winds during the Adrian life cycle is modeled using the Meso-NH model. Two simulations are run with a horizontal grid of 1000 m (Meso1000) and 200 m (Les200). On land, wind speeds are much more underestimated in Meso1000 than in Les200. Higher resolution provides a better representation of topography, and therefore a better representation of local wind speeds. This is in line with the results of Pantillon et al. (2020), who found that winds recorded during the passage of a low pressure system over Central Europe were underestimated with kilometer resolution, whereas they were more accurately captured with LES. Backward trajectories from areas of maximum wind show consistent characteristics of a low-level airflow on the cold side of the curved warm front. This is considered to be the cold conveyor belt, at the origin of the strong mesoscale winds. This is in line with the findings of Raveh-Rubin and Wernli (2016), who show that surface winds are caused by the CCB in the case of a Mediterranean storm, whereas winds are due to a joint contribution from the CCB and SJ, as shown by Brâncuş et al. (2019) in the case of a windstorm affecting the Black Sea.
While the mesoscale origin of winds in extratropical cyclones is well understood thanks to the conceptual model of conveyor belts, the fine-scale processes responsible for the transport of strong winds to the surface and the formation of wind gusts are currently poorly understood. This is because they are not properly represented neither by observation networks nor by numerical weather prediction models. The period of low pressure intensification of Adrian is marked by a combination of thermal instability resulting from a strong air–sea temperature contrast, and strong vertical wind shear leading to dynamic instability. These conditions were favorable for the formation of the convective roll structures detected in the Les200 simulations, below 350-m altitude. Such structures were also found in other meteorological situations [cold-air outbreak (Brümmer 1999) and hurricanes (Foster 2005)] under dynamic conditions very different from those of Mediterranean windstorms. Rolls are recognized for their ability to transport momentum, thanks to their descending and ascending branches (Weckwerth et al. 1997). Previous studies suggested that rolls contribute half (Glendening 1996) or considerably more (Morrison et al. 2005) to the total momentum flux, leading to a significantly increased transfer in the presence of rolls.
The wind is stronger along the downdrafts due to the nonlocal transport of higher momentum air from the CCB through the rolls. In contrast, the wind is weaker along the updrafts that transport low momentum air upward from the surface. These rolls are almost oriented in the mean wind direction with a characteristic size of 2400 m (i.e., 1200 m for each branch) and an aspect ratio of 3. This ratio can vary depending on the situation considered. Rolls are simulated in a hurricane situations with ratios between 1.8 and 6.6 (Nakanishi and Niino 2012), and between 1.5 and 12.7 in cold-air-outbreak situations (Chen et al. 2019). As the windstorm approaches land, convective rolls cannot persist and dissipate rapidly when they reach coastal regions. This suggests that several factors must act on the rolls, such as heating and turbulent mixing induced by the inhomogeneities and greater roughness of the terrain. Strong winds over land are assumed to be transported by orographic forcing, followed by turbulence in the surface layer.
The Meso1000 simulation was unable to represent convective rolls. Using the same atmospheric model and a horizontal resolution of 1000 m, Rivière et al. (2020) found convective rolls in an idealized sting jet case with wavelengths between 5 and 20 km, much larger than the 0.5–3 km observed in a real storm by Browning et al. (2015). This suggests that the convective instability here is not strong enough to produce rolls. Vertical momentum transport was excessively strong compared to the transport in LES. Furthermore, the boundary layer schemes used in mesoscale simulations may not be sufficient to simulate the impact of rolls. This calls for the development of appropriate parameterizations between grid scale winds and subgrid-scale turbulence.
The use of an LES approach then appears necessary to resolve the organized structures and their impact on vertical momentum transport. In this context, two simulations are performed with horizontal resolutions of 100 m (Les100) and 50 m (Les50) to examine whether the horizontal grid spacing impacts the structures found in the Les200. In the area of strong winds, the boundary layer is also organized in convective rolls, as in Les200, despite small differences related to smaller scale structures allowed by the finer resolution. From a spectral analysis, a horizontal grid spacing of 200 m is found to reach an effective resolution of 800 m and to be sufficient to resolve the convective rolls. This result is consistent with previous studies of Huang et al. (2009) who demonstrated that a grid spacing of 250 m was necessary to represent boundary layer rolls in a microphysical study of small cumulus clouds and Thurston et al. (2016) who showed that a horizontal grid spacing of less than 600 m was necessary to model rolls observed under severe fire weather conditions.
Heat fluxes are found to be an important factor in the formation of convective rolls. However, the representation of turbulent fluxes at the air–sea interface in atmospheric models is very uncertain, particularly under strong wind conditions (Pianezze and Barthe 2019). Further research is needed to investigate how the representation of turbulent fluxes influence the presence and the organization of these convective rolls. Although ocean–atmosphere coupling is not relevant for this case study due to the short duration of the windstorm and an almost constant sea surface temperature (not shown), the use of a coupled wave-atmosphere model would also be of interest to examine the potential impact of waves on air–sea fluxes. An obvious limitation of this work is the lack of fine-scale observations, such as synthetic-aperture radar (SAR) observations that provide high-resolution (∼10–100 m) imagery over sea to reveal small-scale structures, such as convective rolls (Foster 2005; Zhang et al. 2008; Brilouet et al. 2023), or wind profilers that provide a better understanding of downward momentum transport over land (Parton et al. 2010). In addition, insufficient availability of buoys has hindered the validation of simulated wind speeds over the sea. Altogether, this emphasizes the great need for measurement campaigns and advanced instrumentation to enable subkilometer-scale observations. To our knowledge, there are very few studies on the fine-scale processes behind the transport of strong winds to the surface in extratropical cyclones. This work could pave the way for a better understanding of the potentially important role of convective rolls and for improving operational forecasts to better account nonlocal vertical transport in strong wind situations, particularly in windstorms causing widespread damages.
Acknowledgments.
Computer resources for running Meso-NH were allocated by CALMIP through Project P20024 and GENCI through Project 0111437. This work was supported by the French National Research Agency under Grant ANR-21-CE01-0002 and by COST Action CA19109. The authors thank Carlos Granero Belinchon and Pierre-Etienne Brilouet for providing the autocorrelation tools and for discussions that contributed to the elaboration of a part of the paper, Météo-France for making the in situ observations available online through the MeteoNet application https://meteonet.umr-cnrm.fr/. The authors also thank Juan Escobar of the Meso-NH support team for his assistance during this study, as well as three reviewers for their constructive comments that helped improve the manuscript.
Data availability statement.
All data are available from the authors upon request.
REFERENCES
Atkinson, B. W., and J. Wu Zhang, 1996: Mesoscale shallow convection in the atmosphere. Rev. Geophys., 34, 403–431, https://doi.org/10.1029/96RG02623.
Bjerknes, J., and H. Solberg, 1922: Life Cycle of Cyclones and the Polar Front Theory of Atmospheric Circulation. Geofysiske Publikationer, 18 pp.
Bougeault, P., and P. Lacarrère, 1989: Parameterization of orography-induced turbulence in a mesobeta-scale model. Mon. Wea. Rev., 117, 1872–1890, https://doi.org/10.1175/1520-0493(1989)117<1872:POOITI>2.0.CO;2.
Brâncuş, M., D. M. Schultz, B. Antonescu, C. Dearden, and S. Ştefan, 2019: Origin of strong winds in an explosive Mediterranean extratropical cyclone. Mon. Wea. Rev., 147, 3649–3671, https://doi.org/10.1175/MWR-D-19-0009.1.
Brilouet, P.-E., P. Durand, and G. Canut, 2017: The marine atmospheric boundary layer under strong wind conditions: Organized turbulence structure and flux estimates by airborne measurements. J. Geophys. Res. Atmos., 122, 2115–2130, https://doi.org/10.1002/2016JD025960.
Brilouet, P.-E., D. Bouniol, F. Couvreux, A. Ayet, C. Granero-Belinchon, M. Lothon, and A. Mouche, 2023: Trade wind boundary layer turbulence and shallow precipitating convection: New insights combining SAR images, satellite brightness temperature, and airborne in situ measurements. Geophys. Res. Lett., 50, e2022GL102180, https://doi.org/10.1029/2022GL102180.
Browning, K. A., 1971: Radar measurements of air motion near fronts. Weather, 26, 320–340, https://doi.org/10.1002/j.1477-8696.1971.tb04211.x.
Browning, K. A., 1990: Organization of clouds and precipitation in extratropical cyclones. Extratropical Cyclones, C. W. Newton and E. O. Holopainen, Eds., Amer. Meteor. Soc., 129–153, https://doi.org/10.1007/978-1-944970-33-8_8.
Browning, K. A., 2004: The sting at the end of the tail: Damaging winds associated with extratropical cyclones. Quart. J. Roy. Meteor. Soc., 130, 375–399, https://doi.org/10.1256/qj.02.143.
Browning, K. A., D. J. Smart, M. R. Clark, and A. J. Illingworth, 2015: The role of evaporating showers in the transfer of sting-jet momentum to the surface. Quart. J. Roy. Meteor. Soc., 141, 2956–2971, https://doi.org/10.1002/qj.2581.
Brümmer, B., 1999: Roll and cell convection in wintertime arctic cold-air outbreaks. J. Atmos. Sci., 56, 2613–2636, https://doi.org/10.1175/1520-0469(1999)056<2613:RACCIW>2.0.CO;2.
Carlson, T. N., 1980: Airflow through midlatitude cyclones and the comma cloud pattern. Mon. Wea. Rev., 108, 1498–1509, https://doi.org/10.1175/1520-0493(1980)108<1498:ATMCAT>2.0.CO;2.
Cavaleri, L., and Coauthors, 2019: The October 29, 2018 storm in Northern Italy—An exceptional event and its modeling. Prog. Oceanogr., 178, 102178, https://doi.org/10.1016/j.pocean.2019.102178.
Chaboureau, J.-P., and Coauthors, 2008: A midlatitude precipitating cloud database validated with satellite observations. J. Appl. Meteor. Climatol., 47, 1337–1353, https://doi.org/10.1175/2007JAMC1731.1.
Chen, Y., H. Yuan, and S. Gao, 2019: A high-resolution simulation of roll convection over the Yellow Sea during a cold air outbreak. J. Geophys. Res. Atmos., 124, 10 608–10 625, https://doi.org/10.1029/2019JD030968.
Clark, P. A., K. A. Browning, and C. Wang, 2005: The sting at the end of the tail: Model diagnostics of fine-scale three-dimensional structure of the cloud head. Quart. J. Roy. Meteor. Soc., 131, 2263–2292, https://doi.org/10.1256/qj.04.36.
Colella, P., and P. R. Woodward, 1984: The piecewise parabolic method (PPM) for gas-dynamical simulations. J. Comput. Phys., 54, 174–201, https://doi.org/10.1016/0021-9991(84)90143-8.
Cuxart, J., P. Bougeault, and J.-L. Redelsperger, 2000: A turbulence scheme allowing for mesoscale and large-eddy simulations. Quart. J. Roy. Meteor. Soc., 126, 1–30, https://doi.org/10.1002/qj.49712656202.
Davolio, S., S. D. Fera, S. Laviola, M. M. Miglietta, and V. Levizzani, 2020: Heavy precipitation over Italy from the Mediterranean storm “Vaia” in October 2018: Assessing the role of an atmospheric river. Mon. Wea. Rev., 148, 3571–3588, https://doi.org/10.1175/MWR-D-20-0021.1.
Deardorff, J. W., 1980: Cloud top entrainment instability. J. Atmos. Sci., 37, 131–147, https://doi.org/10.1175/1520-0469(1980)037<0131:CTEI>2.0.CO;2.
Etling, D., and R. A. Brown, 1993: Roll vortices in the planetary boundary layer: A review. Bound.-Layer Meteor., 65, 215–248, https://doi.org/10.1007/BF00705527.
Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and J. B. Edson, 2003: Bulk parameterization of air–sea fluxes: Updates and verification for the COARE algorithm. J. Climate, 16, 571–591, https://doi.org/10.1175/1520-0442(2003)016<0571:BPOASF>2.0.CO;2.
Flaounas, E., V. Kotroni, K. Lagouvardos, and I. Flaounas, 2014: CycloTRACK (v1.0)—Tracking winter extratropical cyclones based on relative vorticity: Sensitivity to data filtering and other relevant parameters. Geosci. Model Dev., 7, 1841–1853, https://doi.org/10.5194/gmd-7-1841-2014.
Flaounas, E., and Coauthors, 2022: Mediterranean cyclones: Current knowledge and open questions on dynamics, prediction, climatology and impacts. Wea. Climate Dyn., 3, 173–208, https://doi.org/10.5194/wcd-3-173-2022.
Forzieri, G., and Coauthors, 2020: A spatially explicit database of wind disturbances in European forests over the period 2000–2018. Earth Syst. Sci. Data, 12, 257–276, https://doi.org/10.5194/essd-12-257-2020.
Foster, R. C., 2005: Why rolls are prevalent in the hurricane boundary layer. J. Atmos. Sci., 62, 2647–2661, https://doi.org/10.1175/JAS3475.1.
Fouquart, Y., and B. Bonnel, 1986: Computations of solar heating of the Earth’s atmosphere—A new parametrization. Beitr. Phys. Atmos., 53, 35–62.
Friederichs, P., M. Göber, S. Wahl, A. Lenz, and R. Krampitz, 2009: A probabilistic analysis of wind gusts using extreme value statistics. Meteor. Z., 18, 615–629, https://doi.org/10.1127/0941-2948/2009/0413.
Gheusi, F., and J. Stein, 2002: Lagrangian description of airflows using Eulerian passive tracers. Quart. J. Roy. Meteor. Soc., 128, 337–360, https://doi.org/10.1256/00359000260498914.
Giovannini, L., S. Davolio, M. Zaramella, D. Zardi, and M. Borga, 2021: Multi-model convection-resolving simulations of the October 2018 Vaia storm over Northeastern Italy. Atmos. Res., 253, 105455, https://doi.org/10.1016/j.atmosres.2021.105455.
Glendening, J. W., 1996: Lineal eddy features under strong shear conditions. J. Atmos. Sci., 53, 3430–3449, https://doi.org/10.1175/1520-0469(1996)053<3430:LEFUSS>2.0.CO;2.
Granero Belinchon, C., S. G. Roux, N. B. Garnier, P. Tandeo, B. Chapron, and A. Mouche, 2022: Two-dimensional structure functions for characterizing convective rolls in the marine atmospheric boundary layer from Sentinel-1 SAR images. Remote Sens. Lett., 13, 946–957, https://doi.org/10.1080/2150704X.2022.2112107.
Hewson, T. D., and U. Neu, 2015: Cyclones, windstorms and the IMILAST project. Tellus, 67A, 27128, https://doi.org/10.3402/tellusa.v67.27128.
Honnert, R., V. Masson, and F. Couvreux, 2011: A diagnostic for evaluating the representation of turbulence in atmospheric models at the kilometric scale. J. Atmos. Sci., 68, 3112–3131, https://doi.org/10.1175/JAS-D-11-061.1.
Huang, Q., J. H. Marsham, D. J. Parker, W. Tian, and T. Weckwerth, 2009: A comparison of roll and nonroll convection and the subsequent deepening moist convection: An LEM case study based on SCMS data. Mon. Wea. Rev., 137, 350–365, https://doi.org/10.1175/2008MWR2450.1.
Lac, C., and Coauthors, 2018: Overview of the Meso-NH model version 5.4 and its applications. Geosci. Model Dev., 11, 1929–1969, https://doi.org/10.5194/gmd-11-1929-2018.
Lohou, F., A. Druilhet, and B. Campistron, 1998: Spatial and temporal characteristics of horizontal rolls and cells in the atmospheric boundary layer based on radar and in situ observations. Bound.-Layer Meteor., 89, 407–444, https://doi.org/10.1023/A:1001791408470.
Lohou, F., A. Druilhet, B. Campistron, J. L. Redelsperger, and F. Saïd, 2000: Numerical study of the impact of coherent structures on vertical transfers in the atmospheric boundary layer. Bound.-Layer Meteor., 97, 361–383, https://doi.org/10.1023/A:1002641728075.
Lorsolo, S., J. L. Schroeder, P. Dodge, and F. Marks Jr., 2008: An observational study of hurricane boundary layer small-scale coherent structures. Mon. Wea. Rev., 136, 2871–2893, https://doi.org/10.1175/2008MWR2273.1.
Ludwig, P., J. G. Pinto, S. A. Hoepp, A. H. Fink, and S. L. Gray, 2015: Secondary cyclogenesis along an occluded front leading to damaging wind gusts: Windstorm Kyrill, January 2007. Mon. Wea. Rev., 143, 1417–1437, https://doi.org/10.1175/MWR-D-14-00304.1.
Martínez-Alvarado, O., H. L. Baker, S. L. Gray, J. Methven, and R. S. Plant, 2014: Distinguishing the cold conveyor belt and sting jet airstreams in an intense extratropical cyclone. Mon. Wea. Rev., 142, 2571–2595, https://doi.org/10.1175/MWR-D-13-00348.1.
Masson, V., and Coauthors, 2013: The SURFEXv7.2 land and ocean surface platform for coupled or offline simulation of Earth surface variables and fluxes. Geosci. Model Dev., 6, 929–960, https://doi.org/10.5194/gmd-6-929-2013.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, https://doi.org/10.1029/97JD00237.
Morrison, I., S. Businger, F. Marks, P. Dodge, and J. A. Businger, 2005: An observational case for the prevalence of roll vortices in the hurricane boundary layer. J. Atmos. Sci., 62, 2662–2673, https://doi.org/10.1175/JAS3508.1.
Nakanishi, M., and H. Niino, 2012: Large-eddy simulation of roll vortices in a hurricane boundary layer. J. Atmos. Sci., 69, 3558–3575, https://doi.org/10.1175/JAS-D-11-0237.1.
Pantillon, F., P. Mascart, J.-P. Chaboureau, C. Lac, J. Escobar, and J. Duron, 2011: Seamless MESO-NH modeling over very large grids. C. R. Mécanique, 339, 136–140, https://doi.org/10.1016/j.crme.2010.12.002.
Pantillon, F., B. Adler, U. Corsmeier, P. Knippertz, A. Wieser, and A. Hansen, 2020: Formation of wind gusts in an extratropical cyclone in light of Doppler lidar observations and large-eddy simulations. Mon. Wea. Rev., 148, 353–375, https://doi.org/10.1175/MWR-D-19-0241.1.
Parton, G., A. Dore, and G. Vaughan, 2010: A climatology of mid-tropospheric mesoscale strong wind events as observed by the MST radar, Aberystwyth. Meteor. Appl., 17, 340–354, https://doi.org/10.1002/met.203.
Pergaud, J., V. Masson, S. Malardel, and F. Couvreux, 2009: A parameterization of dry thermals and shallow cumuli for mesoscale numerical weather prediction. Bound.-Layer Meteor., 132, 83–106, https://doi.org/10.1007/s10546-009-9388-0.
Petterssen, S., 1956: Motion and Motion Systems. Vol. I, Weather Analysis and Forecasting, McGraw-Hill, 428 pp.
Pianezze, J., and C. Barthe, 2019: Rapport de synthèse sur les paramétrisations des flux turbulents par vent fort: ReNovRisk-Cyclones (C3) Action 2 Sous-action 2.2. LACY Tech. Rep. hal-02923749, 25 pp., https://hal.univ-reunion.fr/hal-02923749/file/rapport_renovrisk_flux_oa_sept2019.pdf.
Pinto, J. G., F. Pantillon, P. Ludwig, M.-S. Déroche, G. Leoncini, C. C. Raible, L. C. Shaffrey, and D. B. Stephenson, 2019: From atmospheric dynamics to insurance losses: An interdisciplinary workshop on European storms. Bull. Amer. Meteor. Soc., 100, ES175–ES178, https://doi.org/10.1175/BAMS-D-19-0026.1.
Pinty, J.-P., and P. Jabouille, 1998: A mixed-phase cloud parameterization for use in a mesoscale non-hydrostatic model: Simulations of a squall line and of orographic precipitations. Conf. on Cloud Physics, Everett, WA, Amer. Meteor. Soc., 217–220.
Pope, S. B., 2004: Ten questions concerning the large-eddy simulation of turbulent flows. New J. Phys., 6, 35, https://doi.org/10.1088/1367-2630/6/1/035.
Raveh-Rubin, S., and H. Wernli, 2016: Large-scale wind and precipitation extremes in the Mediterranean: Dynamical aspects of five selected cyclone events. Quart. J. Roy. Meteor. Soc., 142, 3097–3114, https://doi.org/10.1002/qj.2891.
Ricard, D., C. Lac, S. Riette, R. Legrand, and A. Mary, 2013: Kinetic energy spectra characteristics of two convection-permitting limited-area models AROME and Meso-NH. Quart. J. Roy. Meteor. Soc., 139, 1327–1341, https://doi.org/10.1002/qj.2025.
Rivière, G., D. Ricard, and P. Arbogast, 2020: The downward transport of momentum to the surface in idealized sting-jet cyclones. Quart. J. Roy. Meteor. Soc., 146, 1801–1821, https://doi.org/10.1002/qj.3767.
Saunders, R., and Coauthors, 2018: An update on the RTTOV fast radiative transfer model (currently at version 12). Geosci. Model Dev., 11, 2717–2737, https://doi.org/10.5194/gmd-11-2717-2018.
Schultz, D. M., 2001: Reexamining the cold conveyor belt. Mon. Wea. Rev., 129, 2205–2225, https://doi.org/10.1175/1520-0493(2001)129<2205:RTCCB>2.0.CO;2.
Sheridan, P., 2018: Current gust forecasting techniques, developments and challenges. Adv. Sci. Res., 15, 159–172, https://doi.org/10.5194/asr-15-159-2018.
Shestakova, A. A., P. A. Toropov, V. M. Stepanenko, D. E. Sergeev, and I. A. Repina, 2018: Observations and modelling of downslope windstorm in Novorossiysk. Dyn. Atmos. Oceans, 83, 83–99, https://doi.org/10.1016/j.dynatmoce.2018.07.001.
Shu, C.-W., and S. Osher, 1988: Efficient implementation of essentially non-oscillatory shock-capturing schemes. J. Comput. Phys., 77, 439–471, https://doi.org/10.1016/0021-9991(88)90177-5.
Skamarock, W. C., 2004: Evaluating mesoscale NWP models using kinetic energy spectra. Mon. Wea. Rev., 132, 3019–3032, https://doi.org/10.1175/MWR2830.1.
Slater, T. P., D. M. Schultz, and G. Vaughan, 2017: Near-surface strong winds in a marine extratropical cyclone: Acceleration of the winds and the importance of surface fluxes. Quart. J. Roy. Meteor. Soc., 143, 321–332, https://doi.org/10.1002/qj.2924.
Thurston, W., R. J. B. Fawcett, K. J. Tory, and J. D. Kepert, 2016: Simulating boundary-layer rolls with a numerical weather prediction model. Quart. J. Roy. Meteor. Soc., 142, 211–223, https://doi.org/10.1002/qj.2646.
Weckwerth, T. M., J. W. Wilson, R. M. Wakimoto, and N. A. Crook, 1997: Horizontal convective rolls: Determining the environmental conditions supporting their existence and characteristics. Mon. Wea. Rev., 125, 505–526, https://doi.org/10.1175/1520-0493(1997)125<0505:HCRDTE>2.0.CO;2.
Weckwerth, T. M., T. W. Horst, and J. W. Wilson, 1999: An observational study of the evolution of horizontal convective rolls. Mon. Wea. Rev., 127, 2160–2179, https://doi.org/10.1175/1520-0493(1999)127<2160:AOSOTE>2.0.CO;2.
WMO, 2018: Guide to instruments and methods of observation. WMO/TD-8, 133 pp., https://community.wmo.int/en/activity-areas/imop/wmo-no_8.
Young, G. S., D. A. R. Kristovich, M. R. Hjelmfelt, and R. C. Foster, 2002: Rolls, streets, waves, and more: A review of quasi-two-dimensional structures in the atmospheric boundary layer. Bull. Amer. Meteor. Soc., 83, 997–1002, https://doi.org/10.1175/1520-0477(2002)083<0997:RSWAMA>2.3.CO;2.
Zhang, J. A., K. B. Katsaros, P. G. Black, S. Lehner, J. R. French, and W. M. Drennan, 2008: Effects of roll vortices on turbulent fluxes in the hurricane boundary layer. Bound.-Layer Meteor., 128, 173–189, https://doi.org/10.1007/s10546-008-9281-2.
