Since the development of the Norwegian model of cyclogenesis by Bjerknes (1919) much has previously been written on the flows within mid-latitude cyclones, covering such features as warm conveyor belts (e.g. a 15 year climatology is given by Eckhardt et al., 2004), cold conveyor belts (re-examined by Schultz, 2001) and dry intrusions (e.g. Carlson, 1980). This work is summarized by Browning (1990) in the conveyor belt paradigm. However, this earlier work did not focus on the most damaging winds seen within European windstorms, which are more localized in nature.
Surface wind gust analysis of the 1987 storm lead Browning (2004) to identify distinct regions of the most damaging winds, attributing these to various mesoscale features, with the most damaging winds attributable to a descending mesoscale wind feature at the tip of the cloud head called a sting jet. Clark et al. (2005) extended Browning's observational study with high-resolution numerical modelling, providing much supporting evidence for such a feature and hypothesizing that similar features exist in other storms. Direct evidence for the presence of a sting jet at the tip of the cloud head of a bent back warm front was recently presented by Parton et al. (2009), using observations from two wind profilers during the passage of Windstorm Jeanette across the United Kingdom on 27 October 2002. The profilers captured the descent of the sting jet, with the observations linked through numerical modelling similar to that of Clark et al. (2005).
These latest studies indicate the potential impact of descending mesoscale strong wind features on surface winds and, thus, it is pertinent to examine the mid-tropospheric wind field to determine the frequency of occurrence and nature of such mesoscale features. This work presents a climatology of such features as determined from wind profiler data from the Natural Environment Research Council (NERC)'s Mesosphere-Stratosphere-Troposphere (MST) Radar, located near Aberystwyth, placing the features found in their synoptic context. A feature-detecting algorithm is applied to 7 years' worth of MST radar data to find mesoscale strong-wind regions in the lower and mid-troposphere. Since the radar measures a time series of vertical profiles the term ‘mesoscale’ is used here to refer to the duration of the event—at least 1 h. These features are then categorized with respect to the Norwegian and Shapiro–Keyser (Shapiro and Keyser, 1990) models of cyclogenesis before a brief examination of the nine sting jet cases identified by the study is presented. Finally, through comparison with radiosonde data the wider applicability to the rest of the British Isles is examined.
2. Data requirements and the MST radar
Due to the transitory nature of mesoscale wind features, any data set used to examine features in the free troposphere that may descend to influence surface winds must satisfy the following six criteria:
1.good temporal and spatial continuity;
2.high vertical resolution from the boundary layer top to at least an altitude of 4–5 km;
3.high time resolution to ensure that the general and extreme characteristics of events are adequately resolved;
4.coverage over an entire year to account for seasonality;
5.duration of several years to compensate for annual variation and to give a statistically meaningful climatology, and,
6.consistent quality and continuity despite variations in weather conditions.
Examining the available data for a climatological study of the UK's upper air wind field against the above criteria, only two datasets remain as possible candidates: radiosonde data from the Met Office's network of radiosondes and data from the wind-profiling radars located around the United Kingdom. While the network coverage and high vertical resolution of radiosonde data are arguably of great use, these data suffer from their infrequency of release (typically only every 12 h) and the inability to launch radiosondes in the strongest surface wind conditions. These two points alone limit their applicability to a study of strong wind features that evolve over sub-synoptic timescales and can pass over the United Kingdom in a few hours.
In comparison, wind profilers have, in addition to high vertical resolution (meeting criterion 2), a temporal resolution of a few minutes, meaning that they easily meet criteria 1 and 3, while their reliability and, where these are used operationally, continual use, result in data that satisfies criteria 4 and 5. Their ability to give adequate height coverage (criterion 2) and to have a consistent response in all weather conditions (criterion 6) depends on the type of wind profiler: UHF profilers operating at around 1000 MHz only give limited coverage above 2 km and are subject to contamination by precipitation. VHF radars, however, routinely sense up to and beyond the tropopause and, due to their wavelength being much greater than hydrometeor size, are little, if any, affected by precipitation. Thus, data from VHF profilers are ideal for this study. Within the United Kingdom there are only two such profilers: the NERC's MST radar located at Capel Dewi, Aberystwyth and the Met Office's South Uist VHF profiler. The latter instrument was only installed and operated routinely from mid 2004 onwards and thus does not yet meet criterion 6, while the MST radar has been run quasi-continuously since late 1997 (defined as less than 2% downtime over the course of a year (Eastment et al., 2010)). For this reason, it is the data from the MST wind profiler that form the basis of this study.
The MST radar (see Vaughan, 2002, for full details) operates at 46.5 MHz, using the Doppler Beam Swinging technique to derive the three dimensional wind field above the site by using a combination of Doppler velocities taken along one vertical and two pairs of non-coplanar off-vertical beams. The beam swinging cycle takes around 4 min to complete, giving two wind estimates per cycle. The radar measures winds from 2 to ∼16 km at 150 m intervals, although the pulse length of 2 µs corresponds to a true vertical resolution of 300 m. Although first commissioned in 1990, the radar was not operated continuously until late 1997, and here the 7 years 1998–2005 are used to derive a climatology. There has been a slow deterioration in the sensitivity of the radar since then due to ageing of components, and it is not considered that adding further years to this study would be beneficial (the radar is due to be refurbished in 2010 to recover the original sensitivity).
Although a variety of processing schemes have been used to derive winds from the raw radar data since its initial use, resulting in various versions of the data, the ‘version 0’ data as described by Hooper et al. (2008) have been used for the present study. This product was chosen as it was known to be consistent across the 7 years worth of data and was used to derive the winds used for assimilation into the UK Met Office's Unified Model until recently superseded by ‘version 3’ data: furthermore it did not, at the time of the study, suffer from effects that other processing schemes encountered. However, for ease of use quick look plots from the ‘version 2’ processing were used when cataloguing the strong wind mesoscale events identified by the feature-finding algorithm (for algorithm details see Appendix B of Parton, 2007): where problems and gaps in the data could be quickly identified by eye such events were removed from the climatology. Both versions of the data were obtained from the MST radar data archive held at the British Atmospheric Data Centre (Natural Environment Research Council, 2006)
3. Data analysis
The unaveraged Cartesian wind data available from the version 0 processing, derived from the radial information from three of the non-coplanar beams, gave wind profiles roughly every 2 min (the actual interval varies according to the radar dwell cycle). Within a vertical profile it is possible to have missing or erroneous values in individual height gates. To remove these, median averages were taken from a group of adjacent height gates to obtain a value for the group, hereafter referred to as a ‘level’. Two types of groupings for these levels were used—the first covered the first 20 height gates, grouping them into five height gates per level (750 m). Each of these levels overlapped with the levels above and below by two height gates to ensure redundancy in case one level was still affected by bad or missing data. The second type of level covered the next 80 gates in the profile, grouped to give deeper, non-overlapping layers representative of differing parts of the upper troposphere and lower stratosphere—i.e. level 7 is primarily the free troposphere, level 8 covers the jet stream, level 9 the tropopause region, and level 10 the lowermost stratosphere where gravity wave activity is often seen within the MST radar data (Vaughan and Worthington, 2007). Data above the 10th level were not studied due to their sporadic and noisy nature. Details of all 10 levels are covered in Table I.
Table I. Details of altitude range and height bins used for each ‘level’
Altitude range (km)
Height bins used for level
Various steps were taken to smooth the time series at each level to allow mesoscale structures to be isolated from the highly variable raw winds. First, the data were interpolated to a regular 2 min grid in time, then a median filter was applied with a width equivalent to 10 min to remove outliers. Smoothing to isolate the mesoscale structures in the data was achieved by the application of a Savitzky–Golay filter (see pages 644–649 of Press et al., 2003, for details), applied to preserve peak characteristics such as magnitudes and shapes, but remove higher-frequency variance. (Had a moving window average been used, peak and trough amplitudes would have been suppressed, biasing the results to lower peak wind speeds.) Application of the Savitzky–Golay filter does result in less smoothing of the broader features within the data, but given the need to retain peak wind speeds, and the degree of variability within the data this was deemed an acceptable trade-off.
Following the guidelines (Press et al., 2003) that the filter window should, for a fourth order polynomial fitting scheme, be between one and two times the full width at half the maximum of the features of interest, the window width of the Savitzky–Golay filter was set to be 46 cycles (i.e. 1 h, 32 min) where possible. Occasionally, gaps in the data resulted in continuous periods of data being less than 46 cycles. For these periods, which were still greater than 1 h (i.e. at least mesoscale time periods), the window width was reduced to half of the data period's duration. However, the Savitzky–Golay filter cannot be applied to data that has a width less than half of its window size, resulting in the wings of each data period, typically the first and last 23 min, not being smoothed. To correct for this, a box-car averaging scheme was applied to the data within these wings with a width a quarter of that used for the Savitzky–Golay filter. The smoothed time series were then processed to produce a probability density function of the horizontal wind speeds, along with the corresponding wind roses for each of the levels.
4. Wind rose and wind speed probability density function
Figure 1 shows the probability density function (PDF) for winds on each of the first six levels, together with the average PDF for these levels. From this a number of points are clearly seen.
The mode of each PDF remains between 10 and 12 m s−1. However, the mean wind speeds increase with height from 13.0 m s−1, through 13.3, 14.2, 14.8, 15.9 and 16.8 m s−1 for levels 1–6 respectively. The mean of the average PDF is 13.6 m s−1.
The high speed tail of the distributions become more significant with increasing height.
The probability density drops to zero below 2 m s−1. This is an artefact of the spectral processing in the version 0 scheme: ground clutter returns in the three central points of the Doppler spectra (i.e. a Doppler shift of zero and 1 point either side) are removed by setting those points to zero in the spectra. Since there is always a finite spectral width to the atmospheric echo, this process means the first moment of the spectra (from which the wind is derived) is always non-zero.
There is a small indication of a bimodal structure to the PDFs. Comparison of the PDFs for each year indicate that this is likely to be attributed to annual variability (for example, the PDF for 1998 has a peak which is ∼2 m s−1 faster than the average, but the tail of the distribution at higher wind speed was comparable to that of other years). An alternative explanation is that the PDF is sensitive to the effect of weather systems passing over and the differing distributions that would result by examining such cyclonic and non-cyclonic periods independently. However, this has not been pursued in this study as the focus is on the winds in the tails of the distributions, well away from the bimodal features.
The PDFs in Figure 2 are for the upper levels (levels 7–10). The long tail of the distributions becomes less prominent in the lowermost stratosphere, but the first three levels are fairly similar. The noisy character of these PDFs highlights the variability in features passing over the MST radar site and the wind speeds associated with them: within a day's worth of data winds at these altitudes can vary from 80 m s−1 down to the background wind speed of 10–20 m s−1 as an upper level jet passes over.
At lower tropospheric altitudes, Figure 3, the prevailing wind direction is around 270°, with most winds arriving from the quadrant between northwest and southwest, with a slight bias to the north. Within the upper troposphere and lower stratosphere (levels 7–10 in Figure 4) the wind roses become increasingly narrow as the winds become prevailingly more westerly with height, although a secondary node becomes more pronounced with height at 320°.
5. Identifying strong wind events
Cumulative probability density functions for the lower 6 levels (Figure 5) were used to determine the threshold wind speeds required to constitute a strong mesoscale wind event. These threshold values were used to trial a feature-finding algorithm on the data for 1998. Initially, wind speeds corresponding to the 95th percentile were used for the lowest 6 levels, returning 73 events from the 1998 data, but the resulting estimate of ∼490 events from the entire sample was deemed too large for subsequent analysis. However, trialling the algorithm with the threshold wind speeds set at the 99th percentile returned only 11 events for the 1998 data, which would have lead to only ∼77 events to subsequently examine. While a threshold between the 95th and 99th percentiles could have been used, the importance of the peak wind speeds within such events was introduced as an additional criterion to identify the most damaging events. Thus, a mesoscale wind feature was one which had wind speeds at or above the 95th percentile value for at least 1 h (establishing its mesoscale nature), with peak wind speeds at the 99th percentile value for at least 15 min (to identify the most damaging events). Finally, the feature then had to be present on at least three of the six levels to establish it as a coherent, mesoscale feature. When this modified algorithm was trialled on the 1998 data 19 events were returned, leading to an estimated 133 events within the entire 7 year dataset.
6. Categorization of strong wind events
Application of the feature-finding algorithm to all 7 years' of data returned 130 possible severe wind events for classification. At this point each identified event was examined using the quick-look plots produced routinely by the MST radar facility. These plots provided readily-available and, on the whole, good quality time-height plots of various products of which the following were used: wind speed (with directional arrows every hour), vertical winds, returned signal power, vertical shear of horizontal winds, corrected spectral width and aspect sensitivity. In addition, synoptic charts, satellite imagery and data from the Met Office's Nimrod Rain Radar network (UK Meteorological Office, 2003) were used to give synoptic context. During this stage of the analysis a further 12 events were removed from the sample due to their close proximity to data gaps and one due to known technical problems with the MST radar for that day. No other identifiable data issues were seen for the remaining 117 events. A record of these events is given in Appendix C of Parton (2007).
The classification of each event underwent two stages: first, categorizing the event; second, determining the synoptic placing of the event in relation to an idealized synoptic system. To carry out the first stage the MST radar data and supporting data were examined and compared to the following categories and the characteristics therein, examples of which are shown from profiles on 26 and 27 October 2002 in Figure 6.
6.1. Cold frontal events
Within the radar data a cold frontal event was identified as a sloping feature present in the echo power, wind speed and wind shear plots, starting at low altitudes and increasing in height towards the upper level jet as time increased. In a time-height plot such features are typically seen over a 4–6 h period from 3 to 10 km. Below 3 km the feature often becomes obscured in the echo-power plots by the high background values usually seen below 4 km. A clear example of a cold frontal passage is indicated on the time-height plots by the sloping line between A and B on Figure 6, but becomes hard to trace below 3 km (as a result a frontal fracture in this case). Such events were located, using synoptic charts and satellite imagery, typically on the westward side of the cold frontal cloud band: 32 cold frontal events were found.
6.2. Tropopause folds/warm fronts
Tropopause folds can be traced from the upper level jet down to lower altitudes in the wind speed plots, coinciding with maxima in wind shear and echo-power (Reid and Vaughan, 2004). However, this does not delineate between isolated tropopause folds and those associated with synoptic warm fronts. Both these features exhibit gradients that are typically shallower than cold fronts and they can often be traced from 10 km down to ∼3 km over timescales of 8–16 h. Again, they become hard to distinguish in the echo-power plot below 3 km. An example is seen within the MST radar data in Figure 6 where a warm front is shown by the dashed line on the profiles from 26 October 2002. The supporting information from synoptic charts and satellite imagery helps to distinguish between those events that are warm frontal, found with a close relationship with the cloud band of the following system, and isolated tropopause folds which are found in the clear or shallow cloud zone between successive systems. Twenty six features were found in the sample: the frontal features are distinguished from the isolated folds in the discussion that follows.
6.3. Warm sector events
Events that occurred between an identifiable tropopause-fold/warm-front and subsequent cold front were categorized as warm sector events, labelled ‘A’ on the time-height plots of Figure 6. Later on these will be shown to have characteristics consistent with a warm conveyor belt, but as this is a system-relative feature, such events were categorized as warm sector events at this stage. Supporting information was used to confirm that the 45 events thus classified lay within the system's warm sector.
6.4. Sting jet events
As discussed by Browning (2004), Clark et al. (2005) and Parton et al. (2009), the sting jet is located to the southern side of its parent system's low pressure centre and in close proximity to the tip of the characteristic cloud head hook. Such positioning was determined using the supporting data and then compared to the MST radar data where a wind feature was distinct from but occurred within a few hours of the preceding cold front. Such a wind feature also had to be distinct from the background winds and remain coherent over 2–3 h and a depth of 1.5 km. A clear example is labelled ‘B’ in Figure 6, which was the strongest of the nine sting jet events seen within the data sample: this was the case examined in detail by Parton et al. (2009).
6.5. Unclassified events
At times no clear characteristics were evident within the MST radar data and supporting information was either missing or unable to help with the categorizing of the event. Eight such events were left unclassified.
7. Wind roses of severe wind events
Figure 7 shows the wind roses for the events in each classification. Cumulatively, the strong wind events are predominantly southwesterly as the greatest number of events are the cold frontal and warm-sector events which are strongly southwesterly. In contrast, tropopause folds exhibit mainly northwesterly flows with some variation through to west-southwesterly directions, while all the sting jets have a clear westerly component. Examining the distribution of the unclassified events shows that three of the eight events were from the west-southwest direction, leading to an apparent prevailing wind direction. Although the number of events is statistically too small to point to a new type of feature, examination of their synoptic locations reveals that two of the three events from this direction have some similarities. They could not be categorized using the definitions given earlier, being located behind the cold frontal cloud but south of the cloud head, and there was no indication of tropopause folds within the MST radar data. However, they were both found within Stage IV of the Shapiro–Keyser model of cyclogenesis around 10 h after the passing of a sting jet event. The other event from this direction was located at the triple point of a Stage IV system (see Figure 8f) according to the Norwegian model, where frontal features were unclear.
When the wind directions were re-orientated relative to the schematics (see next section) the main patterns noted above were retained (not shown), with the exception of the unclassified category where the three events noted above re-orientated to a southwesterly direction. Finally, of note is that none of the events had an easterly component in wind direction, either in actuality or in the schematic-orientated wind directions.
8. Placing the events in their synoptic context
During the classification of each event its synoptic placement was compared to a series of idealized schematics displaying the stages of cyclogenesis in the standard Norwegian and Shapiro–Keyser models. The actual synoptic placing of the wind feature and its relation to the system's fronts, low pressure centre and dry slot were used to establish the location and orientation of the peak winds relative to the schematic. The advantage of this approach was to allow systems of varying physical size and orientations to be compared. The process of relating the real synoptic placement to the schematic often involved a translation and rotation from the latitude-longitude grid over to the schematic. Applying the same rotation to the radar wind direction (taken to its closest principal compass point—i.e. N, NNE, NE, ENE etc.) allowed a schematic wind direction to be established for each event for later comparison with other events—again taken to the nearest compass point.
Figure 8 shows the schematic placements of the maximum winds for 103 of the events, where their classification is denoted by the various symbols: tropopause folds/warm fronts by stars, warm sector events by squares, cold frontal events by circles, sting jets by pentagons and unclassified events by crosses. For each event the wind speed and direction (orientated for the schematic) are denoted by the wind barb emanating from the event's classification symbol (half and full barbs correspond to 5 and 10 m s−1 respectively while a triangle denotes 25 m s−1). The wind speeds represent a mean over the levels where the wind feature was detected. The remaining 14 events were difficult to locate synoptically, due either to poor supporting information (charts that did not allow ease of discerning the system behaviour and/or issues with satellite and Nimrod rain radar coverage) or that the parent systems were hard to define as one of the stages of cyclogenesis (this applied to the tropopause folds especially). These were made up of two cold frontal events, one warm sector event, seven tropopause folds, and four unclassified events.
The distribution of event type by synoptic Stage is given in Table II. This clearly shows that half of the events were found in Stages II–IV of the Norwegian model, as compared to fewer than 30% according to Stages II–IV of the Shapiro–Keyser model. Just over a third of events were found in Stage IV of the Norwegian model alone. Similarly, the high proportion of events were accounted for by three of the categories: cold frontal, warm sector and tropopause fold/warm front events (101 in total), with the greatest proportion of these, just over 40% of the sample, being attributed to warm sector events. Such a distribution within the sample highlights the rarity of sting jets, accounting for only nine of the 117 events, or 7.7% of the sample. This translates, if the sample gives a good representation of the population, as around one sting jet event each year where winds attributable to that feature are within the top 5% of mesoscale wind speeds for the mid-troposphere and maximum winds in the top 1%.
Table II. Distribution of events by classification in each model stage
Norwegian stage II
Norwegian stage III
Norwegian stage IV
Shapiro–Keyser stage II
Shapiro–Keyser stage III
Shapiro–Keyser stage IV
When the synoptic distribution of the events is examined, further insights may be gained on the locations of the strongest winds. Firstly, all warm sector events were found ahead of and with flows parallel, or close to parallel, to the surface cold front. Secondly, all but two such events were found within the cold frontal cloud. Thirdly, after being concentrated on the southern side of the low in the early stages of cyclogenesis, the warm sector event distribution extended along the length of the cold front in the later stages, with occasional indication of flow over the warm front (i.e. ana-warm frontal flow). Likewise, there was some indication of ana-cold fronts being present in some of the systems within the sample. These were seen where the warm sector events were just ahead of the cold frontal surface, and towards the centre of the system. They were marked by the presence of a sharp, low-level cold frontal boundary indicative of line convection where the system's warm conveyor belt has a system-relative rearward flow, and rises sharply at the cold frontal boundary at this point. This is similar to the detailed case study by Browning et al. (1998) where cross-frontal motions in an ana-cold front were observed by the MST radar and a co-located 915 MHz RADIAN LAP-3000 wind profiler.
When the schematics are viewed in the context of the conveyor belt paradigm the locations of the warm sector events are consistent with the axis of the warm conveyor belt. Such motion, ahead of, parallel to and along the cold frontal surface, rising either over the cold front towards the centre of the system, or over the warm front, leads to ana-cold and ana-warm fronts respectively. With just two possible exceptions, all the warm sector cases were consistent with a warm conveyor belt.
Cold frontal events did not share the same distribution through all stages of cyclogenesis as the warm sector events, with only one such event being seen in Stage II of the Norwegian model. While clustering towards the centre of the system would appear evident from both Stage III schematics, the Norwegian Stage IV schematic (and to a lesser degree the Shapiro–Keyser equivalent) shows a wider distribution along the cold front up to, but not beyond, the system's triple point. Where those events were found towards the centre of the systems these marked ana-cold front events, while those further out are more typically located where kata-cold fronts would be found.
The majority of tropopause fold/warm front events were warm frontal in nature,—as shown by the proximity to the cloud ahead of the warm fronts. Such flows were parallel to the systems' warm fronts. In contrast to this, the fold events that occurred after the cold fronts are isolated tropopause folds, being found in low cloud or cloud-free regions after the passing of the main cold frontal cloud bands. The majority of the tropopause folds that were unplaced were also isolated in their nature, not being found near to warm frontal cloud.
As shown in Table II, the synoptic placement could not be determined for four of the unclassified events. For the two events in the Stage IV of the Shapiro–Keyser model (Figure 8) their synoptic locations would suggest that one was a warm sector event and the other a tropopause fold. However, these remained unclassified as the MST radar data did not reveal the presence of frontal features as required by the classification scheme used. The two events located in Stage IV of the Shapiro–Keyser model were discussed earlier.
Sting jets were only found within Stages III and IV of the Shapiro–Keyser model. Those in Stage III occurred at the tip of the cloud heads, consistent with the inference of Browning (2004), Clark et al. (2005) and Parton et al. (2009). However, the cluster of events in Stage IV is intriguing as these would appear to be found later than expected and their locations place them further back into the cloud head, leading to the following questions.
1.Were these events found in Stage IV of cyclogenesis or possibly at a late Stage III? The boundary between such systems is hard to define and could lead to an incorrect synoptic placement in this regard.
2.Were these Sting Jet events or jets within the cloud head that had not evolved into the descending Sting Jets? Certainly, figure 14 of Browning (2004) indicates a number of airflows within the cloud making up the slantwise circulations within the cloud head. Here slantwise ascending branches flow around from the north-western side of the cloud head through to the southern side of the central low where they become interleaved with descending branches of air. It is possible, therefore, that the peak winds measured by the MST radar were part of the upward branch of the circulations, while the peak winds were either not located in the flows exiting the cloud head or the storm's advection meant that such flows were not observed by the radar.
3.If these were the descending branches of the slantwise circulation, would the events identified evolve into features that would have an impact at the surface? This question is difficult to answer without studies of the systems over the next 6–12 h, by which time they would be located to the east of the radar.
4.Were these flows associated with multiple cloud heads within the parent systems akin to the system with six stacked cloud heads reported by Dixon et al. (2000), while Browning and Field (2004) and Roberts and Forbes (2004) give details of two other systems with two, discontinuous cloud heads.
By comparison of the wind roses for the classified events it is clear that strong wind events fall into three groups: the warm sector events and cold frontal events from the south-west; tropopause folds from the northwest, and sting jets from the west.
9. Sting jet cases
Table III details the nine sting jet cases identified in this study (referred to hereafter by roman numerals) indicating the maximum rate of central pressure deepening within the parent storm, the time and date of the onset of the sting jet winds identified by the algorithm and the approximate duration of the feature, as well as the stage of cyclogenesis reached by the time the feature was over the MST radar site as discussed earlier. The synoptic development of the parent storms indicate that for all but Case VI the systems had undergone a period of rapid deepening of the order of at least 12 hPa over a 12 h, with five of the cases being ‘explosive deepeners’ (Sanders and Guyakum, 1980). The satellite imagery for all these events (not shown) indicate the presence of the distinctive cloud head hook in close proximity to the MST radar at the time of the features passing aloft. A degree of banding was present within each of these cloud heads comparable to those seen in the events covered by Browning (2004) and Parton et al. (2009). However, while the cloud patterns of Cases III, IV, VII and VIII are more classical in their nature making for ease of identifying following the Shapiro–Keyser model of cyclogenesis, the remaining cases were less conventional due in part to these being in the latter stages of cyclogenesis. In addition, the proximity of other systems in Cases III, V, VII and IX complicated the synoptic scale flow, while of note is the proximity of a secondary cloud feature to the southwest of the cloud head within Case III, redolent of that seen in Figure 2 of Parton et al. (2009) for Case VIII.
Table III. Strong wind duration and synoptic information of the nine sting jet cases
Onset of sting jet winds
Duration of strongest winds (h)
Stage of cyclo-genesis
Maximum rate of deepening mb over (X) h
Figures 9–11 show the wind speed, echo power, horizontal wind shear and vertical wind speeds for the nine events indicating the similarities and differences between the events. Despite the evolution and advection of the storm over the MST radar varying from case to case, meaning that differing parts of the storms were sounded each time, all the events showed a coherent wind feature which was at least 5 m s−1 (and usually over 10 m s−1) above the background winds (which were around 20–25 m s−1) within the sting jet region, with peak winds between 5 and 10 m s−1 stronger still. Case VIII exhibited the strongest jet feature, routinely above 35 m s−1 and with peak winds around 50 m s−1. Upward vertical winds of the order of 1–2 m s−1 were also seen coincident with the wind features.
The degree of banding seen within the MST radar echo power and horizontal wind shear varies between each system. Cases II, VII and VIII show the clearest examples, with the other cases evidencing banding only weakly in one or more radar products.
Parton et al. (2009) have already explored the features within the MST radar data of Case VIII showing that the banding seen in wind speeds, echo power and wind shear lay along wet-bulb potential temperature (θw) surfaces, matching slantwise structure within the cloud head itself. This slantwise structure also continued to be present after the passing of the cloud head aloft. Similar banding, including the persistence after the passing of the cloud head, was also seen in Cases II and VII. Indeed, Browning (2004) presents the MST radar echo power data for Case VII as an example of multiple slantwise circulations within a cloud head.
Browning (2004) also presents satellite imagery and rain radar plots for Case III in discussion of cloud head banding, reporting that this was, ‘an extremely severe event. …it produced sustained winds of hurricane force [over the North Sea]’. Given the rate of deepening that this storm went through, the fact that it passed over the site during this evolution and that Clark et al. (2005) and Parton et al. (2009) have shown that sting jets descend (and accelerate during their descent) from the mid-troposphere, it is likely that the wind feature seen from 0900 UTC to 1100 UTC was associated with the main cloud head and had a role to play in the severe winds experienced in the North Sea some hours later. The remainder of the feature, which can be seen extending to much lower altitudes, arriving over the site from 1100 UTC is more closely associated with the secondary cloud feature to the south-west of the main cloud head. It is unclear what role this plays later in the storm, but may have been responsible for sustaining the hurricane force winds over the North Sea. Unlike the other cases, the banding within Case III, only clear within the horizontal wind shear, is seen at the upper and lower boundaries of the wind feature between 0900 and 1100 UTC within upward vertical winds speeds lower than is seen the other cases. However, from 1200 UTC onwards this increases to the levels encountered in the other cases. Browning (2005) carried out a detailed study of all the significant mesoscale structures seen within this event and reported that the shear layers seen below 5 km do not have inertia gravity wave signatures and that slantwise convection may have had a role in organizing the substructure of the cloud head. Likewise, there was no evidence of inertia gravity waves within any of the sting jet region or the wind tails in Case VIII (Parton et al., 2009).
In all cases sting jets are coherent over a depth of at least 1.5 km and, with the exception of Case VI, they are all found below 7 km. The continuation of strong winds seen within the sting jet region of Case VI into the warm frontal feature seen after 2000 UTC, as well as its differing synoptic development highlight this case to be the least likely to be a sting jet from the nine cases identified in this study.
The well defined slantwise features of Cases II, VII and VIII can be used with the wind speed information to estimate the slope of the slantwise structure within the cloud head. Using representative horizontal wind speeds of 32, 35 and 35 m s−1 for the three cases and using sloping features seen between 1700–1800, 1830–1915 and 0430–0515 UTC respectively, the slopes are found to be ∼1/40, ∼1/43 and ∼1/63 for the three cases. These are consistent with the value of 1/50 used by Clark et al. (2005) to demonstrate the vertical gradient of their model run (and that used by Parton et al., 2009) would be sufficient to capture slantwise circulations generated from the release of conditional symmetrical instabilities.
10. Comparison with UK wind field
The wider applicability of the results presented here is examined by comparing the wind speed PDFs and wind roses of Figures 1–4 to similar plots generated from radiosonde data from stations around the British Isles. Following the approach of Dore et al. (2006) standard radiosonde data held at the BADC (UK Meteorological Office, 2006) for the Stornoway, Camborne, Hemsby and Valentia stations (representing the northern, southern, eastern and western extremes of the British Isles respectively) were analysed. The present work extends the study of Dore et al. (2006) to include the free troposphere as well as the boundary layer, adding an extra year's data and including also data from Aberporth to give a direct comparison with the MST radar (Aberporth is located ∼50 km southwest of the MST radar). Station details are given in Table IV.
Table IV. Radiosonde station details and data used in study
Latitude, longitude (°)
To aid comparison with the MST radar data presented earlier, a series of levels was also produced for the radiosonde data. However, due to the available data being provided with pressure as the vertical axis (no height data were provided) the radiosonde levels were based on equivalent pressure bands to the MST radar data levels. These are detailed in Table V.
Table V. Radiosonde pressure ‘level’ details used for radiosonde—MST radar data comparison
Lower pressure (hPa)
Upper pressure (hPa)
Comparing the wind speed PDF and wind rose for the Aberporth radiosonde data (Figures 12 and 13) with Figures 1 and 2 it is apparent that the distributions from the MST radar data are very close to those from the Aberporth radiosonde with the modes for each level lying between 8 and 12 m s−1. The mean wind speeds are ∼1 m s−1 lower than those for each level from the MST radar data at 12.0, 12.5, 13.2, 14.0, 14.5 and 15.2 m s−1 respectively. This can be attributed to the lack of very low wind speeds in the MST radar data as described earlier and the inability of radiosondes to be launched in very strong surface winds. The principal differences are that the wind speed PDF and wind rose from the Aberporth data are generally noisier, resulting from the far smaller number of points in the radiosonde data, but overall the shapes of the distributions match those of the MST radar data. Using radiosonde data from 1998 to 2004 to be comparable with the MST radar data merely results in a noisier wind speed PDF with no discernable change in characteristics.
Comparing the radiosonde and radar wind roses it is clear that while the overall shape is comparable, the prevailing radiosonde wind direction is at 260° on average, with some indication of moving to being marginally more westerly with height. The offset from the MST radar prevailing direction could be attributed to the lower sampling frequency of the radiosondes and the errors present in the radiosonde (around 5°) and the MST radar data.
Taking these differences into account it is still clear that the wind speed PDFs and wind roses from the MST radar data are comparable to those from the Aberporth radiosondes. Thus, the similarities between the average wind speed PDFs and wind roses for the Aberporth, Camborne, Hemsby, Valentia and Stornoway radiosondes (Figures 14 and 15), indicates that the climatology built up from the MST radar data is applicable to the British Isles as a whole. However, there are a few key differences to highlight from the radiosonde data: (1) wind direction lacks the secondary peak at around 320° identified in the MST radar data, with the possible exception of Stornoway; (2) the Stornoway radiosonde station suggests a possible swing in prevailing wind direction towards southwesterlies with increasing latitude, and, (3) there is a slight shift in the wind speed distribution to lower values from west to east, but little change from south to north across the British Isles.
Comparison of the radiosonde distributions for December to February and June to August (not shown) also indicates a seasonal variation, with a shift to stronger winds in winter and weaker winds during summer at all radiosonde stations. The PDFs for June to August fall to nearly zero above 30 ms−1 indicating that the strongest winds are seen during the winter. Out of the strong wind events identified by the MST radar data only two were seen in June and four in May: all the others occurred between October and January.
Analysis of 7 years' worth of wind profiling radar from the NERC MST radar has enabled wind speed PDFs and wind roses for various heights in the mid-troposphere to be generated. These have been used to constrain a feature-finding algorithm to identify those mesoscale features with wind speeds at or above the 95th percentile and with peak winds at or above the 99th percentile. The resultant 117 mesoscale mid-tropospheric strong wind events were categorized depending on their structure within the MST radar data and found to be primarily cold frontal and warm sector in nature. Tropopause folds, both isolated and associated with the warm fronts, were also found within the events, as were nine sting jet cases (although the classification of one of these events is questionable). The warm sector and cold frontal events were found to have a strongly south-southwesterly prevailing wind direction, while tropopause folds came from the northwest; these three feature types were evident within storms in the latter stages of cyclogenesis according to both the Norwegian and Shapiro–Keyser paradigms. Sting jets were seen to be westerly and in close proximity to a characteristic cloud head hooking around the central depression of the system and were only found within Stages III and IV of the Shapiro–Keyser model of cyclogenesis. In addition, the sting jet cases:
were found within storms that underwent a period of rapid deepening of at least 12 hPa in 12 h during cyclogenesis;
were coherent wind features over at least 1.5 km in depth and 6 h in time sounded by the MST radar;
showed evidence of slantwise structure within the sting jet feature which has been shown by Parton et al. (2009) to lie on θw surfaces, and,
displayed a slope of the slantwise structure has of between 1/40 and 1/63, agreeing with the 1/50 used by Clark et al. (2005).
Finally, the applicability of the study to the whole of the British Isles has been demonstrated through comparison between the MST radar data and wind speed PDFs and wind roses from the five radiosonde stations of Aberporth, Camborne, Hemsby, Stornoway and Valentia.
The authors express their thanks to David Hooper for assistance with the MST radar data and for useful remarks and to the Met Office and the British Atmospheric Data Centre for access to the MST radar and Met Office radiosonde data. The MST Radar Facility at Aberystwyth is funded by the UK Natural Environment Research Council. GP was supported a NERC PhD studentship (award number NER/S/A/2002/10427).