Mesoscale cyclonic storms associated with tornadoes and localised wind damage in frontal rainbands

Ten cases are identified of an unusual type of mesoscale windstorm in the United Kingdom between 2006 and 2021. The windstorms, which occurred in association with frontal rainbands, produced wind damage swaths no more than 20–30km wide, but sometimes extending for several hundred kilometres, often with embedded tornadoes. Storm structure and evolution are explored for two cases using radar data, surface observations, a 1.5‐km gridlength model and damage reports obtained from social media platforms and post‐event damage surveys. Findings are summarised in the form of a conceptual model, and some questions are posed for future research.


Introduction
On several occasions in recent years, narrow corridors of wind damage have been observed in the United Kingdom (UK) in association with mesoscale cyclonic storms embedded in cold-frontal or occlusion rainbands.In each case, the damage swaths were no more than 20-30km wide, but they sometimes extended for several hundred kilometres along the track of the storm (Table 1).In most cases, tornadoes also occurred.Whilst the extent and severity of damage varied greatly, there are some striking similarities in the radar-observed structure and evolution of the storms between cases.Apparently, similar storms have been documented elsewhere in the mid-latitudes.For example, Pinto and Belo-Pereira (2020) investigated a mesoscale windstorm in Portugal that produced wind gusts of over 100 knots.The strongest winds were concentrated within a small core on the immediate southwest flank of the mesoscale circulation, near the hooked tip of an associated cloud head.Tochimoto et al. (2019Tochimoto et al. ( , 2022) ) explored a case of a persistent mesoβ-scale 1 vortex in which several fishing boats were capsized in the Tsushima Strait northwest of Fukuoka, Japan, resulting in five fatalities.A high-resolution simulation revealed embedded tornado-like vortices on the western flank of the vortex (Tochimoto et al., 2019).Although the storm differed from the current cases in that it occurred along the warm front of an extratropical cyclone, the embedded vortices in the model appear similar, in terms of their horizontal scale and position within the larger mesoscale circulation, to radar observations of miso-scale vortices in some of the current cases, as will be described subsequently.
The occurrence of several such mesoscale windstorms in the UK in recent years, the severity of the damage in some of these cases and a lack of understanding of the dynamics of these storms suggest that further study would be beneficial.The purpose of this article is to summarise some key findings from recent cases in the UK, using information from damage surveys, observations and model data.It is hoped that this article will provide motivation for future studies into these events: Synoptic and mesoscale setting, Radar-observed storm structure, Case studies, Summary and questions for further research.

Synoptic and mesoscale setting
At least ten damaging mesoβ-scale vortices within frontal rainbands are known to have occurred in the UK between 2006 and 2021 (Table 1).Although the synoptic settings of these events are somewhat variable (Figure 1), all occurred along cold fronts or occlusions crossing the UK from west to east or southwest to northeast.Most events occurred in autumn.Typically, the extratropical cyclone with which the fronts were associated (hereafter 'parent cyclone') was located over or to the north of the UK.However, some events occurred within ~100km of the parent cyclone centre (e.g. Figure 1h) whilst others occurred further along the trailing frontal system, >1000km from the parent cyclone centre (e.g. Figure 1b).Translational velocity of the mesoβ-scale vortices was rapid (≥15ms −1 ) in all but two cases, and particularly large (>20ms −1 ) in the cases producing the longest damage swaths.
Upper-level flow was strong and generally zonal over the Atlantic and northwest Europe, with the development occurring within a region of substantial cyclonic horizontal shear in all but one case, typically near the left exit of a potent mid-to upper-level jet streak.A composite analysis at 300hPa (Figure 2) shows a positive potential vorticity anomaly (hereafter 'PV anomaly') and associated dry air immediately upstream and poleward of the mesoβ-scale vortex, on the cyclonic shear flank of the jet streak.
The evolution typically involves the PV anomaly and dry intrusion (Browning, 1997) approaching and locally overspreading the surface front from the cold-air side (e.g. Figure 3), whereupon a small, concentrated region of surface pressure falls develops along the front, resulting in a mesoscale depression (hereafter 'mesolow') of relative depth ~1-4 hPa (also see fig. 4 of Clark, 2012, andfig. 8 of Young andClark, 2018).The involvement of an upper-level PV anomaly suggests that the storms may be considered examples, albeit on an unusually small spatial scale, of the 'type 2' cyclonic development of Petterssen and Smebye (1971) (also called 'multi-body' cyclogenesis;Semple, 2003), involving interaction between upper-level features and a surface front.This is in contrast to 'type 1' , or 'single-body' , cyclogenesis, which involves only low-level processes, for example, a hydrodynamic instability along the surface front.In some cases (e.g. as suggested by the sequence in Figure 3d-f), the PV anomaly appears to arise from the breakdown of a previously coherent upperlevel PV filament into a series of localised PV maxima owing, perhaps, to the release of horizontal shearing instability.This view of the development may be compared with the vorticity budget analysis of Tochimoto et al. (2019), in which amplification of the mesoβ-scale vortex was achieved largely through stretching, by updrafts, of environmental vertical vorticity associated with the frontal zone within which the vortex formed.

Radar-observed storm structure
The mesoβ-scale vortices listed in Table 1 were recognisable in radar imagery due to the development of a prominent hook-or S-shaped core of intense rainfall close to the mesolow (Figure 4).Although perhaps superficially similar to hook echoes in supercell thunderstorms (e.g.Markowski, 2002), the hooks in these meso-β-scale vortices are generally considerably larger in hori-1 See Table 2 for a definition of this and other spatial scales referred to in the article.zontal scale.Furthermore, as discussed subsequently, modelling results show that updrafts occupy a narrow strip near the leading, inner edge of the hook, unlike in supercells, where the hook extends around the flank of a deep, persistent, rotating updraft.In some cases (e.g. Figure 5), the hook evolved from part of a narrow coldfrontal rainband (NCFR).In other cases, it developed as a more isolated core of intense rainfall within or near the rear edge of a wide frontal rainband, though sometimes the latter was fragmentary and poorly defined.In yet other cases, the hook was already well formed when the storm moved into radar range (i.e. the formative stages were not observed).Radar data and meso-analyses show the hook to occupy the immediate poleward and cold-air (i.e.north, northwest and west) flanks of the mesolow.In all cases, a gradual increase in coverage and intensity of precipitation occurred on the poleward and cold-air flanks of the mesolow, resulting in an expanding shield of heavy rainfall (labelled 'Sh' in Figure 4).In some cases, banding was evident within this region in radar rainfall fields (e.g.Figures 4b,d,f ), and in the associated, small cloud head evident in visible satellite imagery (e.g.Young and Clark, 2018).Model cross-sections (examples of which are presented subsequently; refer forward to Figure 16) reveal a rearward-sloping frontal structure and suggest the rainfall shield is associated with slantwise ascent, as in the classical ana cold front (Browning, 1990).Consistent with this picture, the hookshaped rainfall core is associated with shallow but strong upright convection at the leading edge of the region of slantwise ascent, at the inner edge of the precipitation shield, where the surface front is thermally and kinematically well marked.Conversely, on the equatorward flank of the mesolow, the front acquires a split crosssectional structure as dry, high-PV air aloft overspreads the surface front.Associated decreases in precipitation intensity and coverage culminate, in cases where the wide frontal rainband was formerly rela-  1.The analysis uses ERA5 data (Hersbach et al., 2020).
In each case, the location of the mesoβ-scale vortex is normalized to 52.5°N, 1.5°W (magenta dot), by shifting the ERA5 dataset horizontally according to the difference between this location and the actual observed vortex location at analysis time in each case.The analysis time in each case was selected as the hour when the storm was at its most damaging.tively coherent, in development of a 'weak echo' or 'echo-free' region ('W' in Figure 4).
Where initially present, the NCFR tends to fracture and dissipate here, although shallow, small NCFR segments sometimes persist (e.g. as in Figures 5d-f ).Although most of the above-mentioned radar-observed features were evident in all analysed cases, their spatial scale varied; for example, on 31 October 2021 (Figure 4f ), the precipitation shield and weak echo region were ~100-150km wide, whereas on 20 October 2021 (Figure 4e), they were only 30-50km wide.In the longer lived cases, these features tended to expand slowly.Doppler radar data, where available, reveal a well-marked mesoscale circulation collocated with the mesolow (Figure 6).Particularly strong gradients in radial velocity occur at the inner edge of the hook-shaped rainfall core on the immediate north, northwest and west flanks of the mesolow, indicative of a narrow band of intense horizontal convergence and cyclonic horizontal shear (dashed line in Figure 6). 2 In some cases, miso-scale circulations (hereafter 'misocyclones'), with diameter ~1-4km, develop along this shear zone.Some of these misocyclones spawn small tornadoes (e.g.Clark, 2012, and as discussed below).The presence of a shear zone and embedded misocyclones on the west flank of the mesolow is consistent with the modelling results of Tochimoto et al. (2019), for the mesoβ-scale vortex over the Tsushima Strait on 1 September 2015.
Several of the attributes described above resemble miniature versions of features in synoptic-scale cyclones.For example, the wellmarked, curved surface front on the northwest flank of the mesolow and the expanding cloud area associated with the rainfall shield resemble, respectively, the back-bent front and the cloud head in larger cyclonic storms of the Shapiro-Keyser type (Shapiro and Keyser, 1990), as emphasised in the schematic model of Young and Clark (2018) for the case of 18 May 2015 over south Wales (their fig.13).Together, these results suggest that the Shapiro-Keyser cyclone model, or at least certain aspects of it, may be applicable to cyclones of a much smaller scale that has generally been reported in the literature to date.
The above-mentioned similarities have also led to speculation that the narrow core of intense winds on the immediate equatorward flank of the mesolow, near to the equatorward tip of the miniature back-bent occlusion, may be a small sting jet (for a recent review of sting jets, see Clark and Gray, 2018).Whilst Pinto and Belo-Pereira (2020) provided evidence of a sting jet in the corresponding part of mesoscale windstorm Xola over Portugal on 23 December 2009, it is not known whether sting jets were involved in the cases listed in Table 1.An alternative hypothesis is that the wind maximum results from super-position of rotational and background flow fields (both contributions being additive on the relevant flank of the storm, assuming the storm moves with the background flow field).Further study is therefore required to identify the mechanisms responsible for damaging winds in this part of the mesolow.

Case studies
In the remainder of this article, results are presented from two cases: 17 November 2016 and 31 October 2021.These cases are 2 Radial convergence is generally not evident along this shear zone in the radar scans shown in Figure 6, owing to the radar viewing angle (the local radial being orientated approximately parallel to the shear zone in each case); however, radial convergence was evident in other scans in each case, where the local radial was orientated at a greater angle to the shear zone.

Introduction
Numerous reports of wind damage appeared in social media on 17 November 2016, initially from the Aberystwyth area in mid-Wales and subsequently from other parts of mid-Wales and the north Midlands.A gust of 84 knots was recorded at the Aberystwyth Lifeboat Station at 1034 utc, followed by a gust of 73 knots at RAF Shawbury, Shropshire, at 1140 utc -a new record for the latter station.Road and rail travel were seriously disrupted by fallen trees and other debris.Damage surveys revealed a swath of tree damage ~8km wide extending across much of mid-Wales and the north Midlands (Figures 7 and 8).Occasional, mostly minor, structural damage was also noted (e.g.roof tiles removed, chimney stacks toppled).Although damage was quite widespread within this swath, the severity of damage varied substantially, sometimes over very small distances.In Shropshire and west Staffordshire, much narrower tracks of intense damage near the northwest edge of the general swath suggested the presence of embedded tornadoes.

Dual Doppler analysis
Network radar imagery shows an intense, persistent, hook-shaped rainfall core moving rapidly east-northeast across Wales and England between 1030 and 1400 utc (e.g. Figure 4b).When over mid-Wales, the hook was sampled simultaneously by two Doppler radars, permitting a dual Doppler analysis (Figure 9).At the radar beam height near the storm at this time (~1km above ground level [AGL]), the analysis reveals a core of >40ms −1 winds southeast of the hook tip and a clear cyclonic circulation in the system-relative wind field, centred immediately east of the hook.A strip of strong cyclonic shear vorticity extends eastsoutheast from the hook tip and around the immediate southern flank of the mesolow circulation, at the northern flank of the wind maximum (blue contours in Figure 9).Near the tip of the hook echo, there also exists a smaller region of strong vertical vorticity stretching (the product of vertical vorticity and horizontal convergence).Bearing in mind the east-northeast movement of the storm, the damage over this part of mid-Wales (white circles in Figure 9) was apparently collocated mainly with the shear zone, immediately northwest of the track  of the wind maximum.However, winds at the radar beam height (~1km AGL) may have differed somewhat from those at the surface.

Synthesis of radar data and damage patterns over Shropshire
A closer comparison of single-site radar imagery and the damage patterns further east-northeast along the storm track is given in Figure 10.Within the general swath of wind damage, there were numerous small pockets (<100-m diameter) of severe tree damage, in addition to areas with little or no damage.Some of the pockets had extremely well-defined edges (e.g. a single tree stripped of most branches, or split through the trunk, with adjacent trees apparently untouched; Figures 7a,e,h).The pockets appeared somewhat random in their distribution.In contrast, coherent, narrow swaths of intense wind damage several kilometres in length were discovered at intervals on the northern flank of the general swath (dashed magenta lines in Figure 10).These tracks apparently originated close to the southeast tip of the hookshaped radar echo, as suggested by their positioning near or slightly north of the path followed by the hook tip.Near their western ends, the tracks were especially narrow (<100-m wide) and had particularly well-defined edges (e.g. Figure 7i), suggesting the presence of tornadoes.However, the tracks often broadened and became less clearly defined with eastern extent, sometimes merging imperceptibly into the general damage swath.
Misocyclones were not observed along the shear zone near the hook tip over Shropshire.However, small-scale maxima in line-of-sight velocities or cyclonic azimuthal shear (i.e.gradients in the line-ofsight velocity between adjacent radials at given range) were sometimes collocated with individual damage tracks (e.g. at 1137 utc in Figure 10), suggesting the presence of vortices too small to be properly resolved by the radar.Misocyclones were later observed at the leading edge of the hook by the Ingham Doppler radar, as the storm tracked through Lincolnshire.The orientation and slight cyclonic curvature of the tracks indicate that, after formation, the vortices responsible travelled some distance around the southern flank of the mesolow, away from the hook tip (as observed directly in the 3 November 2009 case; Clark, 2012).Broadening of the damage tracks and the loss of well-defined edges suggests a transi-tion from tornadic to non-tornadic damage, consistent with high-resolution model simulations of misocyclones in which a flanking wind maximum develops on the southern edge of the vortices as they mature, expand and weaken (e.g.Smart and Browning, 2009;Clark et al., 2021).
Three eyewitness accounts shed further light on some of the observed damage patterns.First, observers at Aberystwyth and Capel Dewi described numerous smallscale, shallow vortices advancing rapidly from west to east, lifting tiles from roofs and snapping tree branches, whilst leaving adjacent trees untouched.These were rendered visible by columns of leaves lifted within the vortices.Second, at Shawbury, an eyewitness described a clearly defined vortex of diameter ~6m (20ft) crossing a field.Short-lived vortices like these are a plausible explanation for some of the localised pockets of intense wind damage without clearly defined tracks.These vortices, like much of the general damage swath, were apparently located southeast of the track followed by the equatorward tip of the hook-shaped echo.Since model fields (not shown) indicate substantial updrafts at >1km AGL only near the leading edge of the hook echo (in agreement with the  respectively).Orange shading is the product of the number of damage points and the mean severity of these points (where severity is the T scale rating + 1) over 0.3 × 0.3 km squares covering the whole domain, with darker shades for higher values of this product.Averaging has been applied over three grid squares in the y direction, and seven grid squares in the x direction (the greater averaging distance in the x direction reflects the fact that the system velocity was much larger in the x direction than the y direction).Green-blue shading indicates areas of reflectivity >38dBZ from the Clee Hill radar in 1° elevation angle scans between 1122 and 1147 utc, as labelled.Blue contours indicate areas of approaching radial velocity in the same radar scans (contour interval 4ms −1 ).The minimum contour value plotted at each scan time varies to account for the changes in approaching velocities associated with the changing azimuth of the storm from the radar; in each case, only the largest values of approaching velocity are plotted.The minimum contour reduces by 4ms −1 in each subsequent scan, from 24ms −1 at 1122 utc to 4ms −1 at 1147 utc.Black contours are azimuthal shear in the same scans (contour interval 2 × 10 −3 s −1 starting at 6 × 10 −3 s −1 ).Bold font indicates gusts measured at two Met Office surface stations along the path of the storm.Letters within boxes indicate the location of the photographs shown in Figure 7, with the letter corresponding to the panel number of Figure 7. Magenta dashed lines indicate probable or confirmed tornadoes.
31 October 2021 case, as described below), it appears that these vortices could not have been collocated with substantial, deep updrafts.This, together with the transient nature and very small scale of the vortices, suggests that they were probably shallow, eddy-like whirlwinds, rather than tornadoes, perhaps a characteristic of the strongly sheared flow on the equatorward flank of the mesolow.
Damage surveys in the USA have revealed similarly complex damage patterns, albeit produced by severe thunderstorms in environments of much larger buoyant instability than was present in the current case.For example, Fujita and Wakimoto (1981) found downburst damage on scales ranging from tens of metres to hundreds of kilometres.The smallest documented structures are comparable in scale to the damage pockets in the current case, suggesting that small microbursts are another plausible explanation for this damage.Forbes and Wakimoto (1983) found evidence of numerous vortices and downbursts occurring nearly simultaneously on a wide range of spatial scales within a bowing cluster of thunderstorms.Tornado-like vortices often occurred at the poleward flank of microbursts, where cyclonic horizontal shear was maximised.Forbes and Wakimoto (1983) noted the potential difficulty in classification of damage as tornadic or otherwise, a difficulty that clearly also applies to the present case.

Introduction
Numerous reports of wind damage appeared in social media on the morning of 31 October 2021.Over time, it became apparent that these reports were concentrated within a narrow swath extending from Dorset to Lincolnshire.Several anecdotal reports of tornadoes were also received, especially from parts of Oxfordshire and Northamptonshire.Much disruption was reported to the road and rail network.Damage surveys conducted by members of TORRO and the public  subsequently revealed a swath of widespread non-tornadic wind damage ~20-25km wide and confirmed that several tornadoes had occurred (Holley et al., 2022; Figure 11).As on 17 November 2016, the tornadoes were generally located towards the left (i.e.northwest) flank of the wider damage swath, looking in the direction of advance of the storm.

Mesoscale setting and surface meso-analysis
The storm developed at the left exit of a 300hPa jet streak, as the latter overspread the rear edge of a frontal rainband moving northeastward across the UK (Figure 12a).Satellite water vapour imagery reveals mid-to upper-level dry air cutting across the surface front and rapid growth of a miniature cloud head on the poleward and cold-air flanks of the mesolow (Figures 12b,c).Surface mesoanalyses show a well-marked mesolow along the surface front and an associated cyclonic flow pattern (Figure 13).These latter features were already well developed when the storm crossed the Dorset coast around 0800 utc.Particularly large gradients in wind speed and direction are apparent on the immediate western flank of the mesolow, near the southern tip of a well-defined, persistent, hook-shaped core of intense rainfall, where vertical vorticity, horizontal convergence and vertical vor-Figure 13.Surface mesoanalysis at 0900 utc 31 October 2021, showing MSLP (navy blue contours at 0.5hPa intervals) and 10m AGL mean wind vectors (arrows) overlaid on radar rainfall rate (shading; for colour scale, see Figure 4).Purple contours are mean wind speed (contour interval 2 knots starting at 34 knots).The analysis uses quality-controlled, time-composited observations from weather stations submitting data to the Met Office's Weather Observations Website and Met Office surface station data, over a period of ±30min from analysis time.Time-composited observations are mapped to a 2km grid using bilinear interpolation.Red dots are individual damage points.Inset: Zoomed view of the mesolow showing vertical vorticity (blue contours at intervals of 1 × 10 −3 s −1 starting at 4 × 10 −3 s −1 ) and vertical vorticity stretching (red contours at intervals of 2 × 10 −6 s −2 starting at 7 × 10 −6 s −2 ).Black dots are individual damage points.Other fields are as in the main panel.ticity stretching are maximised (see inset panel of Figure 13).A ~15km wide core of mean wind speed >34 knots is resolved on the southeast flank of the mesolow, near the tip of a weaker, broader, outer rainfall hook.This outer hook was larger and better developed in the current case than on 17 November 2016, though hints of an outer hook are perhaps evident in the 2016 case too (e.g. in the area immediately west of the wind maximum in Figure 9).Measured peak gusts in this region were in the range 60-70 knots (e.g.67 knots at Netheravon, Wiltshire, and 64 knots at Turweston Aerodrome, Buckinghamshire).

Dual Doppler analysis
A dual Doppler analysis at 0931 utc, when the storm was located over north Oxfordshire (Figure 14), provides independent evidence of many of the features in the surface meso-analysis, including a wind maximum on the southeast flank of the mesolow (~40ms −1 at the radar beam height), near the tip of the outer hook, and the strip of large wind gradients at the leading edge of the inner hook, where vertical vorticity, horizontal convergence and vertical vorticity stretching are maximised.As on 17 November 2016, system-relative winds show a clear cyclonic circulation centred immediately east of the equatorward tip of the inner hook.

Damage patterns as related to radar data
The relationship between radar-observed features and damage patterns documented in post-event surveys is clarified in Figure 15(a).Bearing in mind the northnortheast movement of the mesolow, the narrow swathes of intense wind damage (most of which were confirmed as tornadic) again apparently occurred near the path followed by the southeast tip of the inner hook.The more widely scattered damage points comprising the general damage swath correspond to the position and track of the wind maximum near the outer hook tip.Damage points recorded in the general swath likely constitute only a small fraction of the true number of points, since much of this large area was not surveyed in detail.However, surveys that were conducted within the general swath provided evidence, as on 17 November 2016, of embedded pockets of more concentrated damage (e.g. at Ludgershall and Didcot; Figure 13).
A closer view of damage patterns near the inner hook and comparison with Doppler radar data at 0951 utc is presented in Figures 15(b)-(c).A small-scale wavelike pattern is evident in the reflectivity field, due to the presence at this time of two misocyclones along the shear zone collocated with the inner hook, seen as velocity couplets in the radial wind field (circled in Figure 15c).One of the misocyclones produced a small tornado near Stoke Lyne (path length 5.6km, maximum width ~100m and maximum intensity T2 on the International Tornado Intensity Scale; Meaden, 1976).The other misocyclone subsequently produced a cluster of intense wind damage near Brackley.Individual misocyclones could not be traced unambiguously in the radar data, owing to their rapid translational velocities and short lifetimes, and given the 10-min interval between consecutive Doppler radar scans.However, fracturing, wavelike patterns and misocyclone signatures were evident close to the tip of the hook-shaped rainfall echo at many other times during the lifetime of the storm.

Results from the Met Office 1.5km model
Figure 16(a) shows selected fields from the 0900 utc 31 October 2021 run of the 1.5km gridlength version of the Met Office Unified Model (Davies et al., 2005), valid at 1100 utc.Many of the features evident in the radar data and mesoanalysis are seen in the model fields (cf.Figures 4f and 13), including the precipitation hook, precipitation shield, weak echo region, mesolow and strip of large vertical vorticity at the leading edge of the hook.Vertical sections approximately normal to the synoptic-scale front, through three parts of the mesolow, reveal the contrasting frontal structure alluded to in the 'Synoptic and mesoscale setting' section: rearward-sloping in the section passing through the precipitation shield on the northwest flank (Figure 16b), with a broad and deep region of slantwise ascent, as compared to split structure, with forwardrelative flow above the boundary layer and dry air aloft extending almost to the eastern edge of the section, immediately equatorward of the mesolow centre (Figure 16d).A section through the strongest part of the rainfall hook (Figure 16c) shows a narrow updraft collocated with the hook, extending to ~3km AGL (near the 100km distance marker in the section).Vertical velocity reaches a maximum of ~3ms −1 at ~1km AGL.The updrafts are coincident with a strip of large vertical vorticity (blue contour in Figure 16a), horizontal convergence and vertical vorticity stretching (not shown).Vertical vorticity is maximised at 0.75km AGL where values reach ~6 × 10 −3 s −1 .The juxtaposition of strong updraft and strong near-surface horizontal convergence, together with the limited environmental CAPE (Table 1), suggests that the updraft at the hook is associated mainly with horizontal convergence, rather than buoyant instability (i.e. the convection is 'forced' rather than 'free'), as is the case in many NCFRs (e.g.Browning and Harrold, 1970;Matejka et al., 1980;Carbone, 1982;Hobbs and Persson, 1982).However, some of the other cases listed in

Summary and questions for further research
Observations have been presented from several recent cases of an unusual type of damaging mesoscale cyclonic storm associated with cold-frontal and occlusion rainbands in the UK.A schematic showing the main features of these storms, and their relation to the documented damage patterns, is presented in Figure 17.Whilst the results provide some insights into storm structure and evolution, further research is required to improve understanding and to support nowcasting and warning strategy in future events of this type.Some questions that might usefully be explored by future studies include the following: • Can these mesoβ-scale vortices be regarded as miniature extratropical cyclones (e.g. of the Shapiro-Keyser type) or do the dynamics differ from those of larger-scale cyclonic storms?The results of this study have additionally highlighted the value of crowd-sourced damage reports and post-event damage surveys, for establishing details of the damage patterns, which can then be compared with observations to gain additional insights into storm structure and evolution.Such information could potentially be utilised more fully in future, for example, for the verification of high-resolution model output and impact-based weather warnings.12(a).Thanks also to all who undertook information gathering and contributed to the many informative discussions about these events on the TORRO forum and at subsequent TORRO conferences, to the three eyewitnesses who provided information about the 17 November 2016 case and to members of the public who upload their home automatic weather station data to the Met Office WOW website, which made possible the surface mesoanalysis presented in Figure 13.Finally, many thanks to Martin Young for reviewing an earlier draft of this paper and to the two anonymous reviewers, whose helpful comments and suggestions led to improvements in the content and presentation of this paper.We hope that you are able to personally benefit from your membership.If not, let us fix that!Please email or call us on 0118 2080 142 to learn more about your member benefits.

Figure 3 .
Figure 3. Temporal evolution of selected ERA5 reanalysis fields on 31 October 2021.(a-c) Sequence of west-east vertical sections through the frontal system near to the developing mesolow, at the times indicated above each panel, showing the gradual advance of dry, relatively high-PV air at mid-to upper-levels towards and over the position of the surface front: potential temperature (dashed contours 2K intervals), PV (bold cyan-mauve contours at 1 PVU intervals, starting at 1 PVU) and relative humidity (shading, as per the colour scale to the right of the panels).Red dot denotes the location of the mesolow centre or nascent mesolow centre in each section.'D' indicates the main dry intrusion and associated region of relatively large PV at mid-to upper-levels.Sections are averaged over 3° latitude (13 latitudinal grid points) in each case, centred ~100km south of the developing mesolow.(d-f) Horizontal fields at the times corresponding to the vertical sections in panels (a-c): 300hPa PV (shading), 700hPa relative humidity (dashed contours; magenta = 90%, orange = 60%, yellow = 30%) and 975hPa geopotential height (black contours at 2 DAM intervals).Red dots denote the location of the developing mesolow in each case.'L' denotes the centre of the parent cyclone.Bold blue line indicates the surface cold front and occlusion.Bold, black line and red box indicate, respectively, the central position and full latitudinal extent of the latitudinally averaged sections shown in the corresponding panels (a-c).

Figure 4 .
Figure 4. UK network radar rainfall rate illustrating rainfall features for six of the cases listed in Table 1.Date and time (utc) are indicated above each panel.Annotations are as follows (see main text for further explanation): 'Sh' = stratiform precipitation shield; 'H' = hook echo (or S-shaped echo in the case of panel (c), where frontal fracture on the southern flank of the mesolow has not yet occurred); 'W' = rainfall weak echo or echo-free region; 'SCF' = surface cold front.The black circle indicates the approximate location and extent of the mesolow and associated surface wind circulation in each case.

Figure 6 .
Figure 6.Doppler radar imagery for three mesolow cases, on the dates and times indicated at the top of each column.Top row: radar reflectivity factor; Middle row: radial velocity; Bottom row: zimuthal shear.Radar data are from the Chenies, Clee Hill and Dean Hill radars on 3 November 2009, 17 November 2016 and 31 October 2021, respectively, and the lowest elevation angle scan (1°) is used in each case.The radar is located northeast of the mesolow on 3 November 2009 and south of the mesolow in the other two cases.The area shown in each panel has width 114km.The circulation associated with the mesolow is seen as a local maximum and local minimum in radial velocity on opposite flanks of the mesolow.'L' denotes the approximate centre of the mesolow circulation.Black dashed line in the reflectivity panels indicates the location of the zone of strong cyclonic azimuthal shear on the west or northwest flanks of the mesolow in each case, as seen in the azimuthal shear panels.Positive (negative) radial velocities indicate a component of flow towards (away from) the radar and positive (negative) azimuthal shear is cyclonic (anticyclonic).

Figure 8 .
Figure 8. Overview of the damage swath associated with the 17 November 2016 mesolow.Dashed grey box indicates area shown in Figure 9.

Figure 9 .
Figure9.Dual doppler analysis at 1052 utc 17 November 2016 over mid-Wales using data from the Crug-y-Gorllwyn (Carmarthenshire) and Clee Hill (Shropshire) radars showing, at ~1km AGL, groundrelative wind speed (orange-brown contours at intervals of 2ms −1 starting at 26ms −1 ), storm-relative wind vectors (arrows), vertical vorticity (blue contours at intervals of 0.5 ×10 −2 s −1 ) and vertical vorticity stretching (white contours at intervals of 4 × 10 −5 s −2 , starting at 2 ×10 −5 s −2 ).Underlay is radar reflectivity from the Crug-y-Gorllwyn radar (see Figure6for colour scale).Filled circles denote individual damage locations obtained in ground surveys of the damage by John Mason (TORRO).Wind data are plotted only where reflectivity >10dBZ, since velocities are only retrieved where there is sufficient radar echo.In addition, wind data have been manually masked out where noise broke through in higher reflectivity areas (assessed by inspection of the individual velocity images from which the analysis was constructed); these areas are bounded by the fine white lines near the northwest and southeast corners of the domain shown.Area shown has width 30km.

Figure 10 .
Figure 10.Damage patterns mapped during surveys by the author on 6 days following the storm of 17 November 2016, overlaid on data from the Clee Hill Doppler radar (located ~36km south-southeast of Shrewsbury).Coloured dots are individual documented damage points, with the colour indicating the severity (pink, red and purple equate approximately to damage of intensity T0, T1 and T2 on the International Tornado Intensity [T] Scale,respectively). Orange shading is the product of the number of damage points and the mean severity of these points (where severity is the T scale rating + 1) over 0.3 × 0.3 km squares covering the whole domain, with darker shades for higher values of this product.Averaging has been applied over three grid squares in the y direction, and seven grid squares in the x direction (the greater averaging distance in the x direction reflects the fact that the system velocity was much larger in the x direction than the y direction).Green-blue shading indicates areas of reflectivity >38dBZ from the Clee Hill radar in 1° elevation angle scans between 1122 and 1147 utc, as labelled.Blue contours indicate areas of approaching radial velocity in the same radar scans (contour interval 4ms −1 ).The minimum contour value plotted at each scan time varies to account for the changes in approaching velocities associated with the changing azimuth of the storm from the radar; in each case, only the largest values of approaching velocity are plotted.The minimum contour reduces by 4ms −1 in each subsequent scan, from 24ms −1 at 1122 utc to 4ms −1 at 1147 utc.Black contours are azimuthal shear in the same scans (contour interval 2 × 10 −3 s −1 starting at 6 × 10 −3 s −1 ).Bold font indicates gusts measured at two Met Office surface stations along the path of the storm.Letters within boxes indicate the location of the photographs shown in Figure7, with the letter corresponding to the panel number of Figure7.Magenta dashed lines indicate probable or confirmed tornadoes.

Figure 11 .
Figure 11.Overview of the damage swath associated with the 31 October 2021 mesolow.Adapted from fig. 1 of Holley et al. (2022).Dashed grey box indicates area shown in Figure 14.

Figure 12
Figure 12.(a) ERA5 data showing mean sea-level pressure (MSLP; black contours), 300hPa wind speed (shading) and 300hPa wind barbs (red) at 0900 utc 31 October 2021.Cyan line and 'L' indicate the location of the surface cold front and mesolow respectively.(b) and (c) False-colour METEOSAT Second Generation satellite water vapour images at (b) 0900 utc and (c) 1100 utc 31 October 2021, showing dry intrusion and development of the miniature cloud head.The main features of interest are annotated within each panel.

Figure 14 .
Figure 14.Dual Doppler analysis at 0931 utc 31 October 2021, using data from the Dean Hill and Chenies radars, plotted on Dean Hill radar reflectivity factor.Other contours and vectors are as in Figure 9.

Figure 15
Figure 15.(a) Radar reflectivity factor from the Dean Hill radar at 0949 utc, centred on the mesolow.Various features referred to in the main text are annotated.Damage reports are shown by black dots.(b) and (c) Zoomed-in view of (b) radar reflectivity factor and (c) radial velocity from the Chenies radar at 0951 utc, over an area centred on the inner hook-shaped echo (dashed box in panel (a)).White dashed line in panel (c) indicates the zone of intense horizontal shear (i.e.relative vertical vorticity) along the which misocyclones form.Misocyclones are indicated by black circles.Purple dots are individual damage reports.
Correspondence to: M. R.Clark  matthew.clark@metoffice.gov.uk© 2023 Crown copyright.Weather published by John Wiley & Sons Ltd on behalf of Royal Meteorological Society.This article is published with the permission of the Controller of HMSO and the King's Printer for Scotland.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.doi: 10.1002/wea.4456Renewing your membership is easy!If you have not already, you can renew your membership by direct debit or a one off transaction via your online dashboard.

Table 1
Various attributes of ten damaging meso-β-scale vortices that developed within frontal rainbands in the United Kingdom over the period 2006-2021.Mesolow diameter is estimated from radar data and surface mesoanalyses (where available).Tornadoes are those listed as probable or confirmed in the TORRO tornado database.Damage swath length is the distance between the first and last damage reports along the track and does not necessarily indicate a continuous or near-continuous track of damage over this distance.Location relative to the jet streak indicates cross-axis position (cyclonic shear flank, core, anticyclonic shear flank) and along-axis position (entrance, middle, exit), as analysed using ERA5 reanalysis data.Most-unstable Convective Available Potential Energy (MUCAPE) and 0-1km bulk vertical wind shear are taken from ERA5 reanalysis data, plotted using the thundeR rawinsonde processing tool (available at: www.rawin sonde.com).In each case, a grid-point located on the warm side of the front and immediately south of the mesolow centre is selected, at the hour when the storm was at its most damaging.

Table 2
Definition of horizontal scales referred to in the article.
Figure1.Met Office surface analysis charts for cases listed in Table1.Red dot indicates the location of the mesoscale cyclonic storm at analysis time.

Table 1
had substantially larger CAPE, suggesting a greater contribution from buoyant instability.Weak downdrafts are evident immediately northwest of the rainfall hook, underneath the region of slantwise ascent.
• What factors explain the horizontal scale of development?• What controls the longevity of the mesolow and associated damaging winds?• Can the onset and cessation of damaging winds be anticipated?• What mechanisms are responsible for generation of the wind maximum on the equatorward flank of the mesolow (e.g. are small sting jets involved)?• What does the flow structure look like on the micro-scale within the area of damaging winds and what mechanism is responsible for the localised pockets of severe wind damage?• What determines the balance between tornadic and non-tornadic wind damage, and its variability from case to case? • What is the genesis mechanism of the misocyclones?• Are radar-observed structures of the type described here in always associated with local wind damage or are there 'null' (i.e.non-damaging) cases with similar structures?• Is the development sensitive to other environmental factors such as the amount of CAPE (noting the rather large spread of CAPE values in Table 1)?
Thanks are due to all members of TORRO and the public who conducted voluntary damage surveys, including John Mason, Tim Prosser and Louise Hill following the 17 November 2016 storm, and Sarah Horton, Mary McIntyre, Gordon Robb, Tim Prosser, Louise Hill and Mark Merrony following the 31 October 2021 storm.Thanks are due to Dan Holley for preparation of Figures 8 and 11 and to David Smart for preparation of Figure