Non‐classic extratropical cyclones on Met Office sea‐level pressure charts: double cold and warm fronts

Four cool seasons of Met Office sea‐level pressure analyses revealed 217 extratropical cyclones (<995hPa): 94 had single warm and cold fronts (Norwegian cyclones) and 123 had two cold fronts, two warm fronts, or both two cold and two warm fronts. The most common double‐front cyclone developed when warm and cold fronts were drawn into the circulation of a growing cyclone to produce a cyclone with two cold and two warm fronts. Thus, double‐front cyclones, representing non‐classic cyclone evolutions, are more common than single‐front cyclones in the real atmosphere, and forecasters and scientists should be aware of their existence.

C orrespondence to: Robert L. Wilby r.l.wilby@lboro.ac.uk © 2015Royal Meteorological Society doi:10.1002/wea.2435 however. In the 1980s, field programmes targeting rapidly developing marine extratropical cyclones and mesoscale numerical model simulations showed a slightly different evolution (Shapiro and Keyser, 1990). Instead of a narrowing warm sector during occlusion, as in the Norwegian cyclone model, the Shapiro-Keyser cyclone model exhibited a warm front nearly perpendicular to the cold front in a frontal T-bone, a bent-back front and an eventual warm seclusion (Figure 1(b)). Schultz et al. (1998) and Schultz and Zhang (2007) showed that cyclones in large-scale diffluence tended to form Norwegian cyclones and cyclones in large-scale confluence tended to form Shapiro-Keyser cyclones: the lower friction over water also facilitates the development front, developing single cold and warm fronts, then occluding (Figure 1(a)). First described by Bjerknes (1919) and Bjerknes and Solberg (1922), the Norwegian cyclone model provided a context for scientists and forecasters to describe the evolution of cyclones and led to the modernisation of meteorology in the first half of the twentieth century (e.g. Namias, 1980;1983;Newton and Rodebush Newton, 1994). The model has withstood the test of time, with relatively few modifications. One modification is that of the occlusion process as the wrap-up of the air masses involved rather than the catch-up of the warm front by the cold front (Schultz and Vaughan, 2011).
Not all cyclones undergo an evolution similar to the Norwegian cyclone model,

Katy C. Mulqueen 1 and David M. Schultz 2 Introduction
The evolution of extratropical cyclones has traditionally been described in the context of the Norwegian cyclone model, where a developing cyclone grows on a stationary Non-classic extratropical cyclones on Met Office sea-level pressure charts: double cold and warm fronts of Shapiro-Keyser characteristics (Hines and Mechoso, 1993).
Over the central and eastern USA, other types of cyclone structures and evolutions are possible, including lee troughs and arctic fronts associated with lee cyclones in idealised modelling studies (Hoskins et al., 1984). In the UK, Miles (1962) andBrowning andMonk (1982) recognised the importance of prefrontal troughs and upper-level fronts in the split-front model (also known as a katafront), and multiple fronts have also been documented (Young, 2014).
Consideration of the Met Office sea-level pressure charts for the North Atlantic Ocean and western Europe on a regular basis leads to the realisation that many cyclones possess frontal structures and evolutions inconsistent with the classic Norwegian cyclone model, specifically cyclones with an array of fronts ( Figure 2). For example, the cyclone in Figure 2(a) consists of two warm sectors with two cold and two warm fronts, with a pre-warm-front trough and post-cold-front trough. The Met Office defines a trough as the following: An elongated area of relatively low surface pressure. The troughs marked on weather charts may also represent an area of low thickness (thickness trough), or a perturbation in the upper troposphere (upper trough). All are associated with increasing cloud and risk of precipitation (http://www.metoffice.gov.uk/ guide/weather/symbols). In Figure 2(b) the cyclone is occluded with a bent-back front, an upper cold front, and warm and cold fronts to the north of the cyclone centre. In Figure 2(c), a cyclone with an occluded front, two warm fronts and three cold fronts is connected to an occluded cyclone to its northeast. In Figure 2(d), a cyclone with a bent-back occluded front is associated with three other warm sectors, each with its own cold and warm fronts, and one even having another bent-back front.
Perhaps an initial reaction is to dismiss these 'octopus cyclones' (because of their many radiating fronts from the low centre) as too inconsistent with the Norwegian cyclone model, too inappropriate, too difficult to understand, or unclassifiable (Mass, 1991). Such reactions though reduce the users' confidence in the Met Office sea-level pressure charts as a useful tool for understanding the weather. For example, Prichard (2006) lamented the lack of consistency of features analysed on Met Office charts. Rather than dismiss these analyses, our research is aimed at understanding these non-classic structures taken at face-value from the analyses. As an initial step, this article focuses on a class of cyclones from these charts: those with two warm fronts, two cold fronts, or both two warm fronts and two cold fronts (hereafter all three are classified as double-front cyclones).  (Keshishian et al., 1994;Steenburgh and Mass, 1994) warm-sector rainbands caused by cold fronts aloft (Hobbs et al., 1990;1996;Locatelli et al., 2002) and cyclones with multiple warm fronts (Metz et al., 2004). Multiple fronts have also been documented  Schultz and Vaughan, 2011.) located within the North Atlantic region (40°N-70°N, 50°W-10°E), and was isolated from other cyclones (i.e. no double-barrelled low-pressure centres). This process yielded 217 cyclones in our dataset.
In the second step, the 217 cyclones were classified into either single-front or doublefront cyclones. Single-front cyclones had a single warm front, single cold front, and a possible occluded front throughout their time within the domain described above. Nominally, these could be Norwegian cyclones or Shapiro-Keyser cyclones, or even another type of evolution. Doublefront cyclones had either a double warm front, double cold front, or both, at some point in their evolution within the domain. There were 94 single-front cyclones and 123 double-front cyclones.
The basic structures of these 123 doublefront cyclones are displayed in Figure 3. A double-front cyclone consisted of a cyclone with connected warm and cold fronts passing through the centre of the cyclone (the primary fronts), with an additional front or fronts nearby the cyclone (the secondary front or fronts). The 123 doublefront cyclones consisted of 63 cyclones (51% of the 123) with double warm fronts (Figure 3 In the third selection step we considered the evolution of the double-fronted cyclones. From the 123 cyclones above, a double-front cyclone was further selected if its evolution on the Met Office sea-level pressure charts took place within the domain (40°N-70°N, 50°W-10°E), if the secondary front and the cyclone were less than 200km apart, the secondary front was at least 500km in length and the secondary front was present for at least 6h (appears on two maps 6h apart). As four sea-level pressure charts are produced, at 0000, 0600, 1200 and 1800 UTC each day, 6h is the minimum duration of a secondary front -and hence the minimum period of time during which the cyclone maintained its doublefront structure. (Strictly speaking, a feature that existed on two consecutive maps 6h apart, but not three consecutive maps, could have had a lifetime of at least 6h, but not more than 18h. For the purposes of this paper, we say that a feature appearing on two consecutive 6h maps had a duration of 6h, although we recognise this as a lower limit.) Those cyclones remaining after the third selection step above were also examined for common evolutionary sequences. A double-front cyclone meeting the criteria above was considered classifiable if it was possible to understand how the double front formed based on the evolution displayed on the sea-level pressure charts. from the Met Office global forecast model. In the next step, fronts are initially hand drawn, although automated fronts derived from the 850hPa wet-bulb potential temperature field (Hewson, 1998;Hewson and Titley, 2010) are available to the forecaster to supplement their analyses. The fields referenced by the analyst are partly subjective, but they always include mean sea-level pressure and 850hPa wet-bulb potential temperature. Other commonly used fields include the model's explicitly resolved precipitation rate and parameterised convective-precipitation rate, relative humidity at one or more lower-tropospheric levels, 1000-500hPa thickness and surface dewpoint temperature. The analyst generally looks for discrepancies compared with observations. Where these clearly exist, the analyst will follow the observations not the model fields, but for the vast majority of the time, the model is followed. The analyst is also able to follow the recent history of particular ship and buoy observations, which can be helpful when discrepancies arise.
In this study, we examine the cool-season months October to March because cyclones and fronts are more abundant and better defined in the cool season. The four most recent cool seasons were examined: 2010-2011, 2011-2012, 2012-2013 and 2013-2014. A three-step process was used to identify cyclones. In the first step, we considered just the existence of cyclones. A cyclone was selected for consideration if it had one or more closed isobars, had a central sea-level pressure lower than 995hPa, was studied a cyclone with two cold fronts during the FRONTS 92 field programme. Parker (1998, his figure 1, from T. Hewson) showed a schematic of a frontal wave inside a cyclone's warm sector that yielded a cyclone with two warm and two cold fronts. Schultz (2005) discussed prefrontal troughs and wind shifts, features that could be represented as fronts on surface charts.
The questions we want to examine in this study are how common these non-classic cyclones are and how they evolve. We will examine four cool seasons of Met Office sealevel pressure charts, classifying them as single-front cyclones or double-front cyclones. We will produce a classification scheme for double-front cyclones and describe common structures and evolutions. This study has three goals. First, we aim to provide some guidance for interpreting Met Office sea-level pressure charts with non-classic cyclone structures. Second, we wish to bring a greater appreciation to the variety of cyclone structures and evolutions present in the atmosphere. Third, we want to show that the Met Office sea-level pressure charts are more sensible than they may seem at first from a cursory glance at a single chart.

Data and methods
Our data are the six-hourly sea-level pressure charts for the North Atlantic Ocean and western Europe produced by the UK Met Office and archived at wetter3.de. Production of the charts at the Met Office begins with background fields from a previous model run, typically 6h forecasts

(a) A double warm-front cyclone with a secondary warm front located ahead of the primary warm front. (b) A double warm-front cyclone with a secondary warm front located behind the primary warm front. (c) A double cold-front cyclone with a secondary cold front located ahead of the primary cold front. (d) A double cold-front cyclone with a secondary cold front located behind the primary cold front. (e) A double front cyclone with both a secondary warm front and a secondary cold front.
43 cyclones (81.1% of 53) were classified as category A. Five different evolutions were identified within category A, four of which involved the loss of a front, presumably by merger or frontolysis. Type A1 represents cyclones with two cold fronts where the secondary cold front was located ahead of the primary cold front: 2/53 (3.8%) cyclones formed by this evolution. Type A2 also had two cold fronts, but the secondary cold front was located behind the primary cold front: they were the second most common type of evolution, with 16/53 (30.2%) cyclones. Next, A3 cyclones had two warm fronts where the secondary warm front was located ahead of the primary warm front: they were the most common type of evolution, with 18/53 (34.0%) cyclones. The A4 cyclones also had two warm fronts, but the secondary warm front was located behind the primary warm front: 3/53 (5.7%) cyclones formed by this evolution. Finally, A5 represented doublefront cyclones with both a double warm and a double cold front throughout the evolution of the cyclone: four (7.5%) cyclones underwent this evolution.
Category B cyclones started out as a classic Norwegian cyclone with a single cold front and a single warm front, but with an occluded front to the north that became emplaced on the cyclone, producing either a double cold-or double warm-front cyclone (Figure 7). This evolution is similar to a conceptual model proposed by Metz et al. (2004, their figure 7) for the formation of multiple warm-front-like baroclinic zones within extratropical cyclones east of the Rocky Mountains: attachment of a northern baroclinic zone to a pre-existing cyclone. In our dataset, four cyclones (7.5% of 53) were classified as category B. Type B1 evolved into a cyclone characterised by only a double cold front (2/53, 3.8%), and B2 evolved into a cyclone characterised by only a double warm front (2/53, 3.8%) (Figure 7). Interestingly, this evolution did not produce any cyclones with both a double cold front and a double warm front (although the small sample size of four cyclones could be a factor).
Category C cyclones began life as a cyclone with both a surface warm front Through this process, it became evident that double-front cyclones underwent a small number of classifiable evolutions, allowing for the construction of a classification scheme. A cyclone was considered unclassifiable if its double-front evolution could not be easily recognised from the sea-level pressure charts. Doswell (1991) expressed the importance of having an unclassifiable category, stating that it leaves open the possibility that someone might find a way to classify them in the future. With these additional criteria, the number of classifiable cyclones was 53 (43% of the 123 cyclones) and the number of unclassifiable cyclones was 70 (57%).

Climatology
The distribution of the 123 double-front cyclones shows only a slight preference for some cool-season months over others ( Figure 4). For example, a relative maximum of 26 events (6.5 events per month) occurs in January compared with a relative minimum of 16 events (4 events per month) in November. Given that only four cool seasons were examined, the possibility of the small sample size of the classifiable cyclones producing a different result for a different set of years because of interannual variability precludes further analysis of this graph.  Nevertheless, Figure 4 does show that double-front cyclones on Met Office sea-level pressure charts are relatively common, averaging about five per cool-season month. Indeed, within our dataset of cyclones, double-front cyclones are more commonly analysed than Norwegian cyclones on the Met Office charts.
The distribution of the durations of the secondary fronts in the 53 double-front cyclones is plotted in Figure 5. Although some double-front cyclones existed for only 6h, others existed for as many as 54h ( Figure 5). The mode of the distribution is 12h, with 15 out of 53 cases occurring for this length of time. The mean is 18.7h, and the median is 18h. Thus, double fronts were not isolated features that appeared for a limited period of time only, but were often long-lived and persistently analysed structural features of the cyclones.

Classification
Our classification scheme produced four broad categories of double-front cyclones at some point in their evolution within our analysis domain: A, B, C and D. Category A cyclones started out with both a double cold and double warm front within our analysis domain ( Figure 6): in our dataset,

Discussion
The natural question to ask is where the double-front structure that began the evolution in category A came from. Examination of these events indicated that in many cases the innermost cold and warm fronts to the warm sector resulted from the attachment of an open-wave warm sector at a lower latitude (perhaps associated with the subtropical anticyclone) to a more eastwardly mobile extratropical cyclone at a higher latitude. Figure 10 illustrates one of these evolutions of a cyclone in our dataset classified as type A2. As the mobile cyclone and an upper-level warm front (Figure 8). At some point in the cyclone's evolution, the analyst changed the upper-level warm front to a surface warm front, producing a cyclone with two analysed warm fronts (C1). In our dataset, three cyclones (5.7% of 53) were classified as category C.
Finally, the fourth classification, category D cyclones, had a trough behind the primary cold front. At a later time in the evolution of the cyclone, that trough became analysed as a secondary cold front (Figure 9). In our dataset, three cyclones (5.7% of 53) were classified as category D.
To summarise, the most common evolutions were those in the A classification, specifically types A2 and A3. These two evolutions involve either the weakening of a warm front (A2) or a cold front (A3), leaving a secondary cold or warm front outside of the warm sector (B1, B2, C1, D1). Evolutions leaving fronts within the warm sector (A1, A4 and A5) were relatively uncommon.    passed and deepened from 992 to 960hPa in 24h, it attached to the southernmost 1002hPa open-wave cyclone and became a double cold-front and double warm-front cyclone ( Figure 10). This evolution is similar to the conceptual model proposed by Metz et al. (2004, their figure 6). As the cyclone neared the UK, the leading warm front lost its identity, probably due to a merger with the leading warm front (Figure 10(b) and (c)).
Another question is why these charts had such an abundance of analysed double fronts compared with charts produced by other agencies (National Oceanic and Atmospheric Administration, Deutsche Wetterdienst). We suggest that one reason may be the use of 850hPa wet-bulb potential temperature (θ w ). As gradients in θ w result from gradients in both potential temperature and humidity mixing ratio, some 'fronts' may actually have been moisture gradients. Sanders and Doswell (1995), Sanders (1999), Schultz (2009, p. 355) and Schultz and Blumen (2015) have discussed the importance of analysing fronts as a function of temperature. Although a mask for potential-temperature gradients was available for the forecasters (Hewson and Titley, 2010), comparing some analysed cases with gridded-model output indicated that this might have been a factor for some cases, but not all. Indeed, global frontal climatologies produced using the same θ w scheme have identified 'fronts' in the tropics and subtropics (Berry et al., 2011) that lay along moisture gradients, not temperature gradients (G. Berry, pers. comm.). A more thorough analysis of this issue awaits further investigation.
Although this preliminary study has presented these evolutions and described the most common ones, it does not describe why these double-front cyclones happen. That remains for future research.

Conclusion
The purpose of this study was to examine how common non-classic double-front cyclones are and how they form. Analysing Met Office sea-level pressure charts from four cool seasons between 2010-2011 and 2013-2014 identified 123 doublefront cyclones (<995hPa), exceeding the 94 single-front cyclones (which include cyclones adhering to the Norwegian conceptual model). The 123 double-front cyclones were re-examined, specifically focusing on their evolution. This additional scrutiny led to a dataset of 53 double-front cyclones where the evolution could be classified into one of four categories. Category A (cyclones with two warm fronts and two cold fronts) dominated the dataset with 81.1% of the 53 double-front cyclones. Types A2 and A3 were the most dominant evolutions, with 30.2% and 34.0% of 53 double-front