Investigating cold-frontal gradients in surface parameters using operationally-available minute-resolution data

Authors


ABSTRACT

Large gradients in surface parameters are often observed across cold fronts exhibiting line convection. Such gradients may constitute a hazard to aviation, owing to the associated strong horizontal wind shear and abrupt pressure surges. Significant along-front variability in the magnitude of cross-frontal gradients is usually observed, making quantitative forecasting of expected parameter changes and anticipation of future locations of associated hazards challenging. In this paper, operationally available minute resolution data are analysed for three recent cases of strong cold fronts, in order to explore the magnitude and along-frontal distribution of parameter changes. Cross-frontal temperature fall, sea level pressure rise and wind veer were calculated and mapped. In two of the cases, coherent regions of larger parameter changes were resolved, which could be traced for several hours. This raises the possibility of providing more quantitative short term forecasts of expected parameter changes at downstream locations by extrapolation. Swathes of larger parameter changes usually corresponded to tracks of intense line convection segments. However, significant variability in the relationship between radar reflectivity and the magnitude of parameter changes was found, both within individual cases and between cases. In the third case, the typically small scale of line convection segments precluded the resolution of coherent areas of larger parameter changes, though the general areas in which parameter changes were locally large could still be delineated. Generally, cross-frontal temperature falls exceeding 1.25 °C, pressure surges exceeding 0.4 hPa, and wind veers exceeding 30° could be confidently (>90%) identified by the analysis.

1. Introduction

The UK's location at the eastern end of the North Atlantic storm track ensures that extra-tropical cyclones and their associated frontal systems affect the country frequently, particularly in the winter half of the year. The heaviest rainfall and most abrupt changes in surface meteorological conditions in such systems are typically found at the surface cold front (e.g. Browning and Harrold, 1970; Hobbs and Biswas, 1979). Active cold fronts are frequently characterized by a well-defined, narrow band of heavy precipitation located close to or within the wider frontal rain band, which has been referred to variously as the narrow cold frontal rain band (‘NCFR’) or ‘line convection’. The latter terminology will be used in this paper. Line convection is typically shallow (3–4 km in depth) and is characterized by a narrow zone of intense precipitation, no more than a few kilometres wide, but sometimes extending for hundreds of kilometres in the along-front direction (Browning and Harrold, 1970; James and Browning, 1979). The typical sequence of events at the surface as line convection passes overhead is shown in Figure 1. A rapid temperature decrease (typically of order 1–2 °C) and almost instantaneous pressure increase (typically 0.5–2.0 hPa) occur at, or just before, the time of onset of the heavy (>4 mm h−1) precipitation. An abrupt veer in the surface wind, of order several tens of degrees, and a decrease in speed, are also usually observed. Line convection has long been recognized as a potential source of localized damaging wind gusts and, in some instances, tornadic activity. On occasion, large outbreaks of tornadoes have occurred (e.g. Meaden and Rowe, 1985; Rowe, 1985). For these reasons, line convection may pose a significant risk to various human activities, most notably aviation. Even where winds of damaging intensity are not attained, the strong horizontal wind shear at the leading edge of the line convection may present a significant hazard to aviation, particularly for low-flying aircraft and for those on approach to landing or shortly after take-off (Carpenter, 1976; James and Browning, 1979). The sudden pressure surge occurring at the leading edge of the line also has implications for aviation safety, via its possible effect on the altimeter readings of an aircraft. For these reasons, accurate, location-specific forecasts of the expected magnitudes of wind and pressure changes at frontal passage would be of benefit to the aviation community.

Figure 1.

One-minute observations of temperature, pressure, wind direction and cumulative rainfall at Dishforth, Yorkshire, during line-convection passage on 16 December 2010. This figure is available in colour online at wileyonlinelibrary.com/journal/met

Although line convection is occasionally observed to be highly uniform in the along-line direction over distances exceeding 100 km (e.g. Browning and Pardoe, 1973), substantial along-line variability is usually observed. Typically, the line is organized into precipitation ‘segments’ interspersed by ‘gap’ regions of much lighter precipitation (James and Browning, 1979). A number of different mechanisms have been invoked to explain the segment–gap structure. These include horizontal shearing instability (Wakimoto and Bosart, 2000), the differential motion of individual segments (Wakimoto and Bosart, 2000), trapped gravity waves initiated by strong segments of convection along the line (Brown et al., 1999), waves associated with the vertical environmental shear (Kawashima, 2007, 2011), and the dynamic effects of miso-scale vortices which sometimes form along the leading edge of the line (Trapp and Weisman, 2003). Several of the above studies have demonstrated that the gradients in surface parameters across the front tend to be much larger at the precipitation segments than at gap regions. Observational case studies also reveal that instances of damaging winds and tornadoes tend to occur in association with the more intense line segments (e.g. Hobbs and Persson, 1982). Doppler radar and modelling studies have shown that tornado occurrence is linked to the development of discrete vortices, of typical diameter 1–4 km, within the intense shear zone at the leading edge of the precipitation segments (e.g. Carbone, 1983; Smart and Browning, 2009; Clark, 2011). Although these processes occur on scales well below the average horizontal spacing of the operational surface networks, the larger-scale distribution of cross-frontal gradients in surface parameters might be used to identify, ahead of time, the most intense segments along the line at which these processes are more likely to occur. Operationally, this could aid in directing the forecasters' attention to those parts of a front which most warrant closer examination and monitoring, using other observation types such as Doppler radar. For example, regions of stronger cross-frontal wind veer are likely to be preferred regions of vortex development due to possible vortex sheet ‘roll-up’, resulting from horizontal shearing instability. The magnitude of temperature drop may give an approximate indication of cold pool and downdraft strength, which may also have implications for the likelihood of leading-line vortex-genesis. Since convergence in the cross-frontal direction will tend to sharpen horizontal gradients, a positive correlation between cross-frontal gradients and the low-level convergence, and therefore updraft strength via continuity, would be expected. This likely explains, in part, the spatial correspondence between the strongest frontal gradients and stronger precipitation segments as observed by radar.

Currently, radar reflectivity imagery is the primary tool for determining the most intense (and therefore, by inference, potentially hazardous) sections of a given front in the operational environment. Radar observations lend themselves to this task, since they are of sufficient resolution to observe the vast majority of along-line variability. At present, these data are available at 1 or 2 km resolution over most of the UK, depending on the distance from the nearest network radar. Furthermore, the composite imagery derived from a network of radars provides excellent coverage over a wide area. However, there are limitations. For example, due to the shallow depth of line convection, the radar tends to overshoot the most intense precipitation at greater ranges (e.g. Hobbs and Biswas, 1979). In the UK radar network, the lowest elevation-angle radar beam is typically 1°. Therefore, the beam height would exceed the 3 km lower limit of typical line convection depth at ranges greater than ∼170 km. Despite the reasonably high density of the current UK radar network, a few areas remain (for example, parts of East Anglia) which are located at greater ranges from the nearest radar. Conversely, lines may occasionally be apparent in the radar data, but strong wind gusts and sharp gradients in temperature and pressure are not realized at the surface, for example, owing to the presence of a stable layer at low levels which may prevent the penetration of downdrafts to the surface (e.g. Browning and Roberts, 1999). Furthermore, although a general trend is apparent, the relationship between the observed reflectivity (which is at some range-dependent height above ground) and cross-frontal gradients close to the surface is unlikely to be a simple one, especially given the many factors that can influence radar reflectivity. The surface data therefore have the potential to complement the radar data in active frontal situations by providing a ‘ground truth’, in addition to information about the strength of the front in areas where the radar coverage is poor or non-existent. Additionally, analysis in real time of the cross-frontal gradients and their along-line distribution may allow squall warnings containing more quantitative information, for example about expected wind veer and pressure surges, to be issued to aviation. The success of such a process would depend on the degree of persistence of the larger line elements and gaps, which might allow extrapolation of the observed spatial distribution of parameter changes along the front. In this paper, the potential suitability of the minute-resolution surface data for this task will be tested by analysis of three case studies of line convection. The data and method are described in Section '2. Data sources and method'. Results are presented in Section '3. Results'. Discussion and Conclusions are given in Sections '4. Discussion' and '5. Conclusions', respectively.

2. Data sources and method

The primary data source for this investigation is 1 min resolution surface data. The Met Office has recently completed a major project to modernize the method of surface data collection, storage and dissemination within its network of UK surface stations (Green, 2010). As part of this new Meteorological Monitoring System (MMS), data for all surface parameters are logged at 1 min resolution on site. These data are sent back to the Met Office headquarters in near real-time, where they are stored in a rolling archive for a period of 1 year. The availability of such high resolution surface data, operationally, is a novel aspect of the MMS system. Prior to MMS, stations generally reported at hourly intervals, most commonly in the form of the well-known SYNOP coded messages. Although minute-resolution data were stored locally at each station, they were not available for real-time dissemination and downstream analysis. Surface data relating to each cold front event chosen for analysis was extracted from a rolling archive of data and processed. Unfortunately, over the period in which archived minute data are available, very few cases of intense line convection occurred, owing to a prevalence of anticyclonic conditions with few active depressions crossing the UK. Nevertheless, the chosen cases are the best currently available, and are sufficient to illustrate the potential uses of the minute data, as will be shown. However, none of the cases examined are known to have produced damaging winds or tornadoes.

In order to gain an estimate of the gradients in surface parameters at the front, the maximum change in each of three chosen parameters over a set time-interval was calculated for each station. These values were then plotted manually and contoured. Table 1 shows the parameters of interest, together with details of the calculation of derived parameters. For temperature and sea level pressure, a rolling 5 min difference was calculated for each minute, and the maximum (pressure) and minimum (temperature) values over the period of interest were obtained. The required processing of wind direction data was somewhat more complicated. Large, transient, fluctuations in wind direction, comparable in magnitude to those occurring at the frontal zone, may sometimes occur on a minute-to-minute basis outside of the frontal zone. To filter out the effect of these high frequency fluctuations, a rolling 10 min mean wind speed was calculated for every minute. The 0–360° discontinuity also had to be removed on a case-by-case basis, by using a 180–540° scale where necessary. The maximum wind veer over 10 min was then calculated from the rolling 10 min mean wind direction values. Since abrupt changes in the apparent wind direction can also occur under light wind conditions due to intermittent stalling and movement of the vane, this difference was only analysed for minutes in which the wind speed was ≥ 3 knots. With the exception of isolated examples at particularly sheltered sites, this did not result in the removal of data collected at the actual time of frontal passage, since the mean wind speed was almost always greater than three knots near the line convection, even at low altitude, inland sites. The three knot threshold also ties in well with aviation wind-reporting regulations; large variations in wind direction are not required to be reported when the mean wind speed falls below this value (Civil Aviation Authority, 2008). The effect of extra-frontal variations of other analysed parameters on the analyses produced is discussed in Section '3. Results'.

Table 1. Summary of derived parameters used in the analysis
Parameter (units)Derived parameterConditions
Temperature ( °C)Maximum decrease of 1 min mean temperature over 5 minNone
Sea level pressure (hPa)Maximum increase of 1 min mean pressure over 5 minNone
Wind direction (°)Maximum veer (increase) of the 10 min mean wind direction over 10 minAnalysed only when 1 min mean wind speed ≥ 3.0 knots

The 5 min period over which temperature and pressure differences were calculated was chosen in order to be short enough to minimize the effects of the more gradual changes in parameter values, which might be expected to occur outside of the frontal zone, whilst being long enough to capture the majority, if not all, of the frontal-passage parameter changes. Line convection is rarely more than 3–4 km in width, and the line-normal forward motion usually exceeds 10 m s−1 (James and Browning, 1979). Consequently, the passage of line convection at a given location should not take longer than 400 s (6.7 min) in most cases. In practice, however, the strongest gradients are usually confined to a narrower zone close to the leading edge of the line convection, typically ≤2–3 km in width (Browning and Harrold, 1970; Hobbs and Persson, 1982; Wakimoto and Bosart, 2000). The associated changes in these surface parameters should therefore be largely complete within around 3–5 min. This is generally consistent with surface observations of line convection passage analysed herein, and those presented elsewhere in the literature (e.g. Shepherd et al., 1995; also see Figure 1). A longer period was chosen for evaluation of wind direction changes to compensate for the effect of the 10 min averaging, which will tend to smooth the abrupt change associated with frontal passage.

3. Results

3.1. Case study one: 16 December 2010

A well-marked cold front accompanied by line convection moved south over the UK between 0000 and 1800 UTC 16 December 2010. Figure 2 shows composite radar-rainfall imagery obtained from the Met Office radar network at 4 h intervals, showing the progression of the rain band associated with the cold front. The line convection was widespread and strong over parts of southeast Scotland and northeast England early in the period. Elsewhere, it was weaker and rather more fragmented, with a high degree of along-line variability. As a result, this event provides a good test of the ability of the surface network to resolve the associated along-line variability that would be expected in the cross-frontal gradients of surface parameters.

Figure 2.

Composite rainfall-radar imagery at 0600 (a), 1000 (b) and 1400 UTC (c) on 16 December 2010, showing progression of cold front (‘A’) and post-frontal trough (‘B’)

Figure 3(a) shows the maximum observed 5 min temperature fall over the period 0600–1800 UTC. The radar-observed locations of the stronger line convection segments, at hourly intervals, are also plotted. Overall, there appears to be a good correspondence between the stronger line convection segments and larger temperature falls. Furthermore, the line segments and associated swathes of larger temperature falls appear to be coherent features, which can be tracked for several hours. This is encouraging, since it suggests that short-term (1–3 h), quantitative warnings could be successfully based on extrapolation of recently observed values of temperature fall and its spatial distribution. The segments and gaps apparently moved slowly eastwards along the front as the front itself moved south. This is in agreement with previous observations of line convection which have shown that the segments move along the front in the direction of the pre-frontal low-level flow (e.g. Hobbs and Biswas, 1979). Larger-scale features which were resolved in this case include a swathe of substantial temperature falls extending from northeast Wales and Cheshire through the west Midlands and into central southern England, a swathe of large but decreasing falls extending from northeast England, through the east Midlands and into Cambridgeshire, and a swathe of much smaller temperature falls corresponding to the track of a persistent gap region extending from Cumbria and parts of Lancashire through to southeast England. Figure 3(b) and (c) show similar analyses for maximum 5 min pressure increase and maximum 10 min wind veer, respectively. Comparison with Figure 3(a) shows good agreement overall, with the same larger-scale features being apparent in the analysis of all three parameters. However, some of the smaller-scale detail apparent in the temperature analysis is not apparent in the pressure analysis, for example, owing to the smaller number of stations at which pressure observations are currently available.

Figure 3.

Analyses of (a) maximum 5 min temperature fall, (b) maximum 5 min pressure increase and (c) maximum 10 min wind veer over the period 0600–1700 UTC 16 December 2010. Bold lines show the radar-observed locations of line convection segments at 1 h intervals (alternating solid and dashed). Shaded regions indicate areas of larger parameter changes, according to the key at the bottom of the figure. This figure is available in colour online at wileyonlinelibrary.com/journal/met

Some care is required in the interpretation of values over areas bordering the Irish Sea, since larger gradients (in particular, of temperature) were noted locally in association with post-frontal showers in these areas. This explains the larger values in some locations where line convection appeared to be absent along the cold front, for example on the Isle of Man and at one station in southwest Scotland. The situation is also complicated in this case by the presence of a post-frontal trough over eastern parts of the region of interest, which also exhibited weak line convection at times (see Figure 2). In general, the temperature falls, pressure increases and wind veers associated with this feature were smaller than those associated with the cold front. However, these anomalies illustrate the potential difficulties associated with the analysis, and correct attribution of abrupt parameter changes to frontal passage or post-frontal convection.

In order to investigate this aspect further, the temperature data were re-analysed, in conjunction with radar data, in order to determine whether the maximum temperature fall over the analysis period was actually associated with frontal passage. Figure 4 shows the probability that the maximum observed parameter change over the analysis period occurred at frontal passage, as a function of the magnitude of the parameter change at frontal passage (i.e. the probability that the analysis correctly ‘isolated’ the frontal passage changes). When the temperature fall exceeds 1.5 °C, there is a very high probability (>95%) that the maximum observed fall over the whole period occurred at the time of frontal passage. However, for temperature falls of less than 0.5 °C, there was only a 2% chance that the maximum observed fall over the whole period occurred at the time of frontal passage. This is because 5 min fluctuations in temperature exceeding 0.5 °C frequently occurred outside of the frontal zone (i.e. the ‘signal’ associated with frontal passage was of a similar magnitude as, or smaller than, the ‘noise’ associated with non-frontal temperature fluctuations). The analysis therefore suggests that, in this case, 5 min temperature falls exceeding approximately 1.5 °C can be considered significant (corresponding to the shaded regions in Figure 3(a)). Similar analysis of pressure and wind direction data showed that 5 min pressure increases of ≥ 0.4 hPa and 10 min wind veers of ≥ 40° can confidently (>90%) be attributed to frontal passage (see Figure 4). The shading in Figure 3 corresponds to these limits.

Figure 4.

Probability of the maximum observed parameter change over the analysis period occurring at the time of frontal passage, as a function of the magnitude of the maximum parameter change. This figure is available in colour online at wileyonlinelibrary.com/journal/met

3.2. Case study two: 25 March 2010

A cold front moved northwards over southern England between 1500 and 1900 UTC 25 March 2010. Scattered thunderstorms broke out along the front as it approached the south coast, quickly merging to form a squall line which met mesoscale convective system size criteria (American Meteorological Society, 2000) for at least 3 h as it moved through central-southern and southeast England. Figure 5 shows the surface analysis chart for 1800 UTC and radar data at hourly intervals between 1600 and 1800 UTC. Minute data for the period 1400–2100 UTC were analysed and the maximum differences in temperature, pressure and wind direction were calculated in the same way as described for case study one. Figure 6 shows plotted and contoured parameter differences. The temperature analysis (Figure 6(a)) shows a relatively small swathe within which maximum temperature falls exceeded 1 °C widely, and 2 °C locally. This area corresponds to the track of an intense line segment which moved northwards through central southern England and into the Midlands between 1600 and 1900 UTC (marked ‘A’ in Figure 6(a)). The correspondence between the observed temperature falls and the radar-observed line is particularly good along the sharply-defined western boundary of the segment. A secondary, smaller, region of larger temperature falls is evident further east, extending from Sussex into Essex. This was apparently associated with a separate line segment (‘B’ in Figure 6(a)). Outside of the areas affected by the squall line, maximum observed temperature falls were almost always < 1.0 °C, and in the majority of cases < 0.5 °C.

Figure 5.

(a) Surface analysis at 1800 UTC 25 March 2010. Isobars are drawn at 1 hPa intervals and warm and cold fronts are marked using conventional symbols, and composite radar rainfall imagery at (b) 1600 UTC, (c) 1700 UTC and (d) 1800 UTC 25 March 2010. Rainfall-rate scale as in Figure 2. This figure is available in colour online at wileyonlinelibrary.com/journal/met

Figure 6.

(a) Maximum 5 min temperature fall, 1400–2100 UTC (b) maximum 5 min pressure increase, 1400–2100 UTC (c) maximum 10 min wind veer, 1400–2100 UTC 25 March 2010, and (d) temperature fall (shading as in panel (a)) overlaid on ATDNet lightning data. In (a) and (d), solid lines show locations of line segments at hourly intervals, with times in UTC. Individual line segments are labelled as in the main text. This figure is available in colour online at wileyonlinelibrary.com/journal/met

Figure 6(b) shows plotted and contoured maximum 5 min pressure increases. As in case study one, some of the finer-scale detail apparent in the temperature analysis is not reproduced in the pressure analysis, owing to the lower number of pressure measurements. However, the larger-scale features were reproduced. For example, the swathes of larger pressure increases associated with line segments ‘A’ and ‘B’ are both apparent. Outside of the area affected by the squall line, maximum pressure differences were universally ≤0.4 hPa, suggesting that the ‘limit’ of pressure increases which may be considered significant is very close to that found in case study one.

Figure 6(c) shows plotted and contoured values of maximum 10 min wind veer over the analysis period. In general, the wind veer experienced at locations affected by the squall line was not significantly greater than that observed elsewhere. Maximum values in the area affected by the line are generally in the range 30–50°. The double-maximum structure associated with segments A and B was briefly evident around 1800 UTC, but in other areas affected by the line, including those at which the largest temperature falls and pressure rises occurred, no significant veer was observed. In fact, the largest observed 5 min wind veer over the analysis area occurred over Wales, well outside the region affected by the squall line, with values exceeding 60° along a narrow, south-southwest to north-northeast orientated swathe. Inspection of the surface analysis chart at 1800 UTC (Figure 5(a)) suggests that these large values occurred along the eastern flank of a low pressure centre and close to the triple point of the associated northward moving frontal system, which crossed Wales during the period 1700–2100 UTC.

A further interesting feature, as revealed by the temperature analysis, is that at stations affected by segment ‘A’ in its early stages (e.g. those on the Isle of Wight and along the adjacent mainland coast), rapid temperature falls were not observed as the line passed. This suggests that a significant system-scale cold pool had probably not developed by this time. However, as the line moved inland and matured, significant temperature falls were increasingly observed, suggesting that a substantial cold pool, with a well-defined leading edge, was developing during this period. Lightning location fixes from the Met Office ATD system (Keogh et al., 2006) are shown in Figure 6(d), together with radar-observed locations of the line segments at hourly intervals through the period of interest. In the case of segment A, the lightning occurred mainly along the eastern parts of the line, and to the east of the swathe of maximum observed temperature falls. Therefore, whilst the most active convection apparently occurred along the eastern flanks of the segment, the strongest cold pool (and associated temperature gradients) developed along the western flank where somewhat less intense leading-line convection, but a developing area of trailing stratiform precipitation, can be seen in the radar data. This is likely a result of along-line movement of individual cells towards the west, such that mature and weakening cells with well-developed downdrafts and cold pools tended towards the western parts of the line, whilst new cells developed predominantly towards the eastern end. The lightning activity decreased and eventually ceased altogether as the line segment moved into the Midlands. However, significant temperature decreases (and pressure increases) continued to occur in association with the leading edge of the line for at least another 1–2 h, presumably due to the persistence of the cold pool and gust front for some time after weakening and decay of the line commenced.

3.3. Case study three: 14 September 2010

A cold front moved south across England and Wales on this day. Line convection occurred in some places, though the front was rather fragmented and weak over much of the west of the area (Figure 7). This event, therefore, provides a marginal case, in which the limits of the analysis method described herein can be further explored. The line convection morphology was also rather complex at times, exhibiting a double structure in places (e.g. see Figure 7(b)). Parameter differences were calculated over the period 0400–2100 UTC. As in the 16 December 2010 case, heavy showers moved into northern parts of the area during the latter part of this period (for example, see Figure 7(c)), associated with weak instability in the post-frontal airmass. These showers locally produced temperature, pressure and wind fluctuations exceeding those occurring at the time of frontal passage. For stations within the northern part of the area, the analysis was performed over the whole period, and then repeated using a shorter period ending around 3–4 h after frontal passage at each station, in order to eliminate the effects of the post-frontal showers on the analysis. Results corresponding to analysis with and without the effect of the post-frontal showers were then compared.

Figure 7.

Composite rainfall-radar imagery at (a) 1200 UTC, (b) 1400 UTC and (c) 2000 UTC 14 September 2010. Rainfall-rate scale as in Figure 2

Figure 8 shows maximum observed wind, pressure and temperature differences obtained from the analysis following removal of the effects of the post frontal showers. A complex pattern is revealed by the temperature analysis (Figure 8(a)), with the largest 5 min decreases generally found in a swathe extending from North Wales and southern parts of Northern England, through to the north and east Midlands, and East Anglia. Temperature falls exceeding 2.5–3.0 °C occurred locally within this area. However, substantial localized variations in maximum temperature falls can also be seen. This is probably a result of the marked along-line variability in line convection on scales equal to, and smaller than, that of the average station spacing in this area (for example, see the 1400 UTC radar rainfall image; Figure 7(b)). In contrast to the 16 December 2010 case, there do not appear to be any coherent maxima within the general zone of larger temperature falls that can be tracked for several hours. The situation may be exacerbated in this case by the relatively low density of surface stations over East Anglia. The general zone of largest falls is orientated mainly in the along-front direction, suggesting a rather transient intensification of the frontal zone along a broad swathe; the intense phase appears to have persisted longest in the east. Outside of this area, maximum temperature falls were generally less than 0.5–1.0 °C, though a small area of slightly higher falls is evident along a swathe extending from Somerset towards the Isle of Wight.

Figure 8.

Analyses of (a) maximum 5 min temperature fall, (b) maximum 5 min pressure increase and (c) maximum 10 min wind veer over the period 0400–2100 UTC, 14 September 2010. Lines show radar-observed location of line convection segments at hourly intervals (alternating dashed and solid). This figure is available in colour online at wileyonlinelibrary.com/journal/met

The correspondence between locations of radar-observed line convection and the areas of larger temperature falls is less good, as a whole, in this case than for those previously described. For example, although the line convection elements over the east Midlands produced locally large temperature falls, the more coherent line segment that moved through the west Midlands (e.g. Figure 7(b)) only produced relatively small temperature falls, in spite of its apparent intensity in radar imagery. Again, this illustrates the possible variability in the relationship between radar reflectivity and gradients in surface parameters. Conversely, the analysis reveals areas of larger temperature falls over northwest Wales and northern parts of East Anglia, where no radar-observed line occurred. These areas are located at relatively large ranges from the nearest network radar. The most likely explanation, therefore, is that the radar beam overshot the line convection in these areas. The surface analyses add extra value here, since the magnitude of 5 min temperature decreases suggests a strong likelihood that line convection was present, which would not have been evident using radar data alone.

The pressure analysis (Figure 8(b)) reveals that few stations experienced significant pressure surges at frontal passage (i.e. > 0.5 hPa). A narrow zone of slightly higher values extends from North Wales to East Anglia, which corresponds reasonably well to the area of larger temperature falls. However, pressure surges even within this area were modest, at 0.5–1.0 hPa generally. This may be due, in part, to the smaller number of stations for which pressure observations are available; for example, the two stations in East Anglia that experienced temperature falls exceeding 3 °C have no pressure observations. However, observations at stations having both temperature and pressure observations within this area do suggest that pressure surges were modest in this case, even where quite large temperature falls, and strong radar-observed line convection, occurred. This emphasises the potential complexity and variability in the relationship between the magnitudes of different parameter changes at frontal passage. Outside of the zone of larger pressure increases, values were generally less than 0.5 hPa. However, a small zone of larger values was evident near the south coast, between Somerset and the Isle of Wight, which is in agreement with the small area of larger temperature falls observed here.

Figure 8(c) shows the maximum 10 min wind veer over the analysis period. Comparison with Figure 8(b) reveals excellent agreement in the spatial distribution of areas with the largest wind veers and pressure increases. As for the pressure analysis, the number of stations which experienced a significant wind veer (i.e. > ∼40–50°) was small. Outside of these areas, the observed maximum wind veer was less than 40° at all stations.

Figure 9 shows the probability that the maximum parameter changes over the analysis period occurred at the time of frontal passage, as a function of the magnitude of the parameter change. Chances of detection of the wind veer at frontal passage increased to over 90% as the veering increased to 20° or more, even before removal of the effect of post frontal showers. This is a smaller veer than could be confidently detected in the 16 December 2010 case (see Figure 4). For temperature, the ‘threshold’ frontal decrease for confident detection (>∼90%) was around 1.0 °C before removal of the effect of the showers, but only 0.5 °C if the showers are removed by limiting the analysis time to ∼3 or 4 h after frontal passage at each site. This is also somewhat smaller than was found in the 16 December case. Confident detection of 5 min pressure increases occurred for values ≥ 0.4 hPa, regardless of whether the effect of the showers was removed or not. This is similar to the value obtained in the 16 December case.

Figure 9.

Probability of the maximum observed parameter change over the analysis period occurring at the time of frontal passage, as a function of the magnitude of the maximum parameter change. Dashed lines show probabilities derived from the full analysis period (0400–2100 UTC). Solid lines show probabilities derived following limitation of the analysis period at northern stations in order to remove the effect of post-frontal showers. This figure is available in colour online at wileyonlinelibrary.com/journal/met

4. Discussion

4.1. Potential applicability for short-term forecasting

The ability of the minute data to reveal coherent features in the patterns of parameter changes at frontal passage in two of the three cases analysed suggests that there is some potential for successful short term warning of squall severity at locations ahead of active cold fronts, and possibly linear convective systems more generally. In these cases, quantitative estimates of wind veer, pressure surge and temperature fall could be forecast by extrapolation, based on recently-observed values, out to perhaps 2 or 3 h ahead. Such information could be of benefit to aviation activities in particular, which are sensitive to abrupt changes in winds and pressure. The greatest benefit is likely to be achieved by using the surface observations in conjunction with numerical models. For example, if the data could be successfully assimilated into the models in real time, it might better constrain the distribution and intensity of line convection segments and gaps, which could lead to more realistic short-term evolution of the line in the model. If successful, more accurate, location-specific short-term forecasts of expected parameter changes and their along-front variability could be extracted directly from the model. Unfortunately, it is not currently possible to assimilate surface data of such high temporal resolution into operational numerical models, though investigation into this area is ongoing.

The Met Office currently employs a precipitation nowcasting tool, STEPS, which uses a blend of radar data and model rainfall and wind fields to provide a short-term forecast of the rainfall distribution, and an estimate of the uncertainty in that forecast (Bowler et al., 2006). It may be beneficial to investigate whether nowcasts of cross-frontal parameter changes could be produced using a similar blend of model and observations data. Alternatively, it would be interesting to investigate whether the minute-resolution surface observations could be used to improve precipitation nowcasts in the STEPS system. For example, it could be speculated that better resolution of sharp cross-frontal gradients might improve rainfall nowcasts in line-convection situations, by improving the representation of some of the low-level forcing mechanisms that influence rainfall generation (e.g. low-level convergence at the leading edge of the line and cross-frontal density gradients).

Ultimately, integration of the surface observations with available model data is likely to be the most elegant way of using the additional information provided by the minute data, since the benefit would transfer downstream automatically via the various products which are routinely generated from the model forecasts. This aspect is likely to be a primary objective of any future work on this subject at the Met Office. However, for illustrative purposes, an example of a ‘manually-generated’ forecast of cross-frontal parameter changes is presented here (Figure 10). The forecast was produced by extrapolation, based on previously observed cross-frontal differences, with no consideration of the radar data. In this example, locations in region ‘B’ could be warned to expect 5 min temperature falls exceeding 1.5 °C widely, with falls of 3.0 °C locally. Within the same region, pressure surges of greater than 0.5 hPa, locally > 1.0 hPa, may be expected. Locations within region ‘C’ could be warned to expect pressure surges exceeding 0.5 hPa, with temperature falls exceeding 1.5 °C locally, though with values reducing over time, based on the observed weakening over the previous 6 h. Comparison with Figure 3(a) and (b) suggests that the forecast would have provided reasonably good guidance in this case. However, the success of this method will clearly vary from case to case. Quantitative forecasts of parameter changes based upon extrapolation of current values and longer-term trends clearly will not be as accurate in cases exhibiting rapid development and decay of individual line convection elements (e.g. as observed in case three). Analysis of more events is required before any general conclusions can be made about the potential benefits of this type of analysis. However, the initial results presented herein offer some promise.

Figure 10.

(a) Maximum 5 min temperature falls and (b) maximum 5 min pressure increases over the 6 h ending 1200 UTC 16 December 2010 (shaded regions indicate changes exceeding 1.5 °C and 0.5 hPa). Bold, solid line shows ‘current’ (1200 UTC) location of cold front. Bold, dashed line indicates the expected position of the front 3 h later, based on extrapolation. The hatched areas labelled ‘A’, ‘B’ and ‘C’ highlight regions in which larger temperature and pressure changes may be expected within the next 3 h, based on extrapolation of the observed trends over the preceding 6 h. This figure is available in colour online at wileyonlinelibrary.com/journal/met

For the purposes of this study, it is unfortunate that damaging, widespread line-convection events have been lacking since the archived minute data has become available. Consequently, the potential relationship between areas of large parameter changes and damage reports cannot be explored. It might be expected that the analysis would prove more reliable in stronger line convection cases for providing quantitative forecasts of parameter changes, since frontal gradients should be more clearly distinguishable from the non-frontal gradients occurring during the remainder of the analysis period. Furthermore, in the stronger cases, line convection tends to be more widespread, reducing the issues associated with sampling of small line convection segments, which were typical of the rather fragmented, weak line convection present in the cases analysed herein. Nevertheless, these rather marginal cases have been useful for exploring the limitations of the method. Analysis of future cases will be conducted as they continue to occur.

The analyses presented herein show that there is some case-to-case variability in the ‘threshold’ values of parameter changes that can confidently be attributed to frontal passage. Table 2 shows threshold values for cases one and three and the mean value for the two cases (the equivalent analysis could not be performed for case two owing to the fact that that cold front only traversed part of the analysis area). Analysis of further cases will be required to define these threshold values, and their case-to-case variability, better. In the cases analysed, post-frontal convection appears to be the primary source of extra-frontal parameter changes of magnitudes comparable to those observed at frontal passage. Such effects may be expected in many frontal situations, since showers frequently occur in post-frontal polar maritime air on the western and southern flanks of active depressions that tend to be the most prolific producers of strong cold fronts exhibiting line convection. The effect of post-frontal showers has been investigated here by limiting the period of analysis to 3 or 4 h after frontal passage. Removing the effect of the showers reduces the threshold values. This could be achieved operationally by using analysis tools which show, for example, the maximum parameter changes over the previous 3 or 4 h: this could more effectively ‘isolate’ values associated with frontal passage, whilst retaining enough data to identify the tracks of larger parameter changes in order to extrapolate likely future values successfully. The ideal period of analysis would depend on the line-normal rate of advance of the front and proximity to the front of post-frontal showers, for example. Operational experience suggests that there would usually be sufficient separation between the frontal band and post-frontal showers to allow distinction between the parameter changes associated with each.

Table 2. ‘Threshold’ values of parameter changes for confident (>∼90–95%) detection, before removal of the effect of post-frontal showers
Parameter (units)Case one thresholdCase two thresholdMean threshold
5 min temperature fall ( °C)1.51.01.25
5 min pressure increase (hPa)0.40.40.4
10 min wind veer (°)402030

Parameter changes associated with post-frontal showers may also, in themselves, be of interest to the forecaster, provided that they can be distinguished from changes associated with frontal passage. This is because they may be associated with similar hazards as might be expected with the frontal line convection (as described in Section '1. Introduction'). However, it must be accepted that the larger parameter changes will be highly localized due to the typically small horizontal scales of post-frontal convection. For this reason, coherent patterns would not usually be expected, except in the case of well-defined defined post-frontal troughs. Figure 11, which shows maximum 5 min temperature decreases over the 3 h period ending 1800 UTC on 14 September 2010, illustrates these points. Comparison with radar data (e.g. Figure 7(c)) shows that the patchy areas of large temperature falls over the north of England and north Wales were associated with post frontal showers, whilst the areas of large falls extending from East Anglia towards the west country were associated with the frontal band. The two regions are separated by an ∼200 km-wide swathe of small temperature falls, associated with the gap region between the cold front and post-frontal convection, during the analysis period. Displaying a rolling sequence of similar 3 h periods in a loop may further help to clarify the situation in an operational environment.

Figure 11.

Maximum observed 5 min temperature fall for the 3-h period ending 1800 UTC 14 September 2010. Dashed lines show locations of cold-frontal line convection segments at 1800 UTC. Solid lines show locations of the leading (southeastern-most) post-frontal showers at the same time. This figure is available in colour online at wileyonlinelibrary.com/journal/met

4.2. Other potential applications

Although the analysis herein has been limited to cold-frontal line convection events, a similar mapping of abrupt parameter changes could potentially be used in other types of situation. Non-frontal mesoscale convective systems are perhaps the most obvious example, especially those which exhibit a linear, leading line structure. Such systems often produce flooding and lightning damage, and, somewhat more rarely in the UK, damaging wind gusts and tornadoes. The analysis could be used to confirm whether a system is ‘surface-based’ or elevated, for example. This has a bearing on the associated hazards. For example, the risk of strong wind gusts and tornadoes is much higher when convection is surface-based. As illustrated by case study two, the analysis may also provide some information about the system developmental stage and structure, which would help to provide clues about expected future trends in associated hazard occurrence, magnitude, and spatial distribution. A second potential use of the temperature and wind direction analyses is for identification of sea breeze passages. At present, it is unclear whether the signal associated with sea breeze passage would be large enough to be distinguished from the noise associated with general ‘background’ parameter fluctuations. Future case studies of sea breeze events will be analysed in order to investigate the viability of this idea. Finally, the temperature analysis could be used on radiation nights to monitor the rate of decrease of surface temperatures, which could be particularly beneficial, for example, in the monitoring and early warning of ice and frost risk, or to warn of surface temperatures rapidly approaching fog point.

5. Conclusions

Three cases of cold-frontal line convection have been analysed in order to investigate the applicability of minute-resolution surface data for analysis and short-term forecasting of the magnitude of abrupt parameter changes at frontal passage. The results suggest that such analyses would provide a useful short-term forecasting tool in many cases of cold frontal passages, or indeed non-frontal convective events in which linear morphologies predominate. However, it is evident even from the limited number of cases analysed here that the structure of cold frontal line convection varies substantially from case to case. As a result, the ability of the analysis method to provide meaningful, quantitative, measures of parameter changes and their spatial distribution will also vary. In two cases, stronger line segments were sufficiently large and long-lived to produce coherent features in the analysis, such that short-term forecasting based on extrapolation of recent trends would be possible. Conversely, in one case, the dominant scale of line convection was smaller than the average spacing of the surface stations. As a result, coherent structures were not produced in the analyses of parameter change magnitudes. Nevertheless, it was still possible to identify a broader region within which parameter changes were locally large.

In addition to the ongoing analysis of further line-convection cases, future work will concentrate on automation of the analysis and contouring procedure. This will facilitate analysis of a much larger set of cases, in addition to providing a system that is more suitable for operational use. Archiving of data pertaining to many frontal events should eventually provide a valuable source of data for research into the general structure and dynamics of line convection. Finally, the possibility of integrating the analysis with existing automated nowcasting systems at the Met Office will be explored.

Acknowledgements

The author would like to thank the anonymous reviewers for their helpful comments and suggestions. Thanks are also due to Malcolm Kitchen and Mike Molyneux for revising earlier drafts of this article.

Ancillary