The Numerous Approaches to Tracking Extratropical Cyclones and the Challenges They Present

Walker, E. , Mitchell, D. M., & Seviour, W. J. M. (2020). The numerous approaches to tracking extratropical cyclones and the challenges they present. Weather, 75(11), 336-341. https://doi.org/10.1002/wea.3861

Publisher's PDF, also known as Version of record License (if available): CC BY Link to published version (if available): 10.1002/wea.3861

Link to publication record in Explore Bristol Research PDF-document

This is the final published version of the article (version of record). It first appeared online via Wiley at https://doi.org/10.1002/wea.3861. Please refer to any applicable terms of use of the publisher.

University of Bristol - Explore Bristol Research General rights

This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/red/research-policy/pure/user-guides/ebr-terms/ The numerous approaches to tracking extratropical cyclones and the challenges they present

Erin Walker1, of ETCs (Bengtsson et al., 2006; Michaelis Storm tracks are part of a very com- 1 et al., 2017). plex coupled system with many different Daniel Mitchell and The pathways along which ETCs typi- interacting components that can strongly William Seviour1,2 cally travel are known as storm tracks. influence an ETC’s location and intensity. 1School of Geographical Sciences, Climatological storm-track regions are prev- Changes in the location of storm tracks, alent areas of synoptic-scale disturbances both latitudinally and zonally, have been University of Bristol, UK where, for example, there is a maximum linked to the subtropical jet, baroclinicity 2 Global Systems Institute and polewards transport of energy occurring and extratropical sea-surface temperatures Department of Mathematics, in the North Pacific and North Atlantic (Brayshaw et al., 2009; 2011; Woollings et University of Exeter, UK oceans in the NH (Blackmon, 1976; Booth al., 2010; Feser et al., 2015). In addition, et al., 2017). In the Southern Hemisphere storm tracks respond to large-scale phe- Weather – November 2020, Vol. 75, No. 11 No. 75, Vol. 2020, – November Weather Introduction (SH), during summer, the storm track nomena such as the El Niño Southern forms a circular pattern around Antarctica, Oscillation, the North Atlantic Oscillation Extratropical cyclones (ETCs) are large-scale which becomes more asymmetric in win- (NAO), the Quasi-Biennial Oscillation and low-pressure systems that develop in the ter (Hoskins and Hodges, 2005; Ulbrich the Madden-Julian Oscillation (Ulbrich et mid-latitude regions. These systems can et al., 2009). During winter, baroclinicity is al., 2009; Feser et al., 2015; Yang et al., 2015; travel thousands of kilometres and can last at a maximum in both the Pacific and North Wang et al., 2017, 2018). For example, several days and are often, but not always, Atlantic Ocean basins (Nakamura, 1992; Hurrell et al. (2003) illustrated how storm associated with high winds and heavy rain Hoskins and Hodges, 2019). In terms of track activity and ETC intensity increase in (Ulbrich et al., 2009; Catto, 2018). The most baroclinic wave activity, the North Atlantic regions of the North Atlantic Ocean during powerful ETCs can cause significant socioec- storm track reaches maximum intensity a positive NAO (Figure 1a). In addition, links onomic damage, costing millions of pounds during winter, whereas the Pacific storm have been identified between storm tracks (Hawcroft et al., 2012; Garnier et al., 2018). track has a mid-winter minimum (due and changes in the stratosphere during win- For example, large storm surges associated to the especially strong jet stream), with ter (Kidston et al., 2015). with ETCs can cause loss of life through maximum intensity occurring during late The changes in the position and inten- coastal flooding, while strong winds cause autumn and early spring (Nakamura, 1992). sity of storm tracks will impact the local falling trees and debris, in addition to dis- The SH storm track maximum intensity (i.e. climate and weather over large distances rupting transport systems and severely strongest ETCs) also occurs during winter, (Bengtsson et al., 2006). The North Atlantic damaging property. with enhanced activity in the southern jet stream is eddy driven and therefore con- Cyclogenesis can occur in numerous ways, Atlantic and Indian Ocean regions (Hoskins nected to the North Atlantic storm track. the most ubiquitous of which is baroclinic and Hodges, 2005; Ulbrich et al., 2009; They both normally exhibit a similar south- instability characterised by strong vertical Booth et al., 2017). west–northeast orientation (Figure 1a), wind shear in the mid-latitudes. This shear, in turn, results via thermal wind balance, due to strong temperature gradients. ETCs act to reduce this gradient through the polewards transport of latent and sensible heat. Consequently, if the gradient is small, there is less potential energy available for cyclogenesis. Decreased equator-to-pole temperature gradients in the lower troposphere, resulting from polar amplification (the increased rate of warming in higher latitudes compared to lower latitudes as a result of increasing concentrations of green- house gases (Manabe and Wetherald, 1975)), are believed to be one reason behind the predicted decrease in ETC numbers for the Northern Hemisphere (NH) (Bengtsson et al., 2009; Catto et al., 2011). In addition, the increase in temperatures would enhance Figure 1. Schematic showing the physical processes over the North Atlantic Ocean during latent heat release and is thought to con- both phases of the Arctic Oscillation and the North Atlantic Oscillation. (Figure adapted from tribute to the deepening and intensification Greene, 2012, p 54.) 336 The numerous approaches to tracking extratropical cyclones Weather – November 2020, Vol. 75, No. 11 337 ------., 2006; et al ., 2015).Most ., 2018). ., 2013). Some et al et al . cre (2019) have ., 2019).Others use et al ., 2008; Massey, 2016; ., 2016). Filtering tracks tracks ., 2016). Filtering et al et al et al et al ., 2003; Bengtsson ., 2019). ., 2008;Massey, 2012, Neu2016; ., 2013; Grieger et al et al et al et al ., 2013; Pinto ., 2013; Pinto ., 2008; Massey, 2016). Quite often, Quite tracks Massey, 2016). ., 2008; hours (Hoskins and Hodges, 2002, 2005; Once identified, an ETC must be tracked tracked be must ETC an identified, Once In and tracking identifying addition to and trackingNew identification tech objective feature-tracking methods have objective feature-tracking methods have and ETC of an the identification phases: two multiple across trackingsame system the time-steps (Raible Lakkis known is what rise to giving time, through Tracking problem. as the correspondence an ETC identify algorithms able to must be in the system same that and then identify track point Neighbour time-step. following or minimum value ing uses a local maximum this and then tracks variable of a climate nearest-neigh time using a through point bour model (Lakkis and smoothness functiona cost improve to track same the match points that ensure 2016). In(Hodges, 1994, 1995; Massey, 2012, various con addition, methods implement the possibility of match reduce to straints (Hoskinsing errors and Hodges, 2002), for based radius setting a search instance, (Raible speed of an ETC on the average et al only selected are features that so filtered are 1000km exceeds length track total the if and/or lasts longer than 24, 48 or even 72 Hodges speed and intensity (Feser (Feser intensity and speed maximum point within one level of data data of level one within point maximum However, time. through point that and track 3-dimensional complex features have ETCs in the multiple levels extend that through Lakkisatmosphere. 4-dimensionala feature-tracking ated (4D) ETCs tracks and identifies algorithm that in the atmosphere. multiple levels across the method from adapted have They Raible et al which some standardisation, provide help to multiple studies across can be implemented (Neu it is equally impor time, through an ETC tracked accurately it is that ensure to tant encoun be can that Issues space. through latitude–longitudein changes include tered with increasing decrease that sizes box grid discrepancy), leading to (resolution latitude multi are There singularities the poles. at these problems, address to ple approaches data truncating filters, spatial from ranging re-gridding data a certain at wavenumber, all grid, a different and projecting it onto unique tracking algorithms of which create (Hodges, 1994; Hoskins and Hodges, 2002; Massey, 2012; Zappa of these methods can be computationally only to limited while others are expensive, one hemisphere ETCs track being able to a time. at more capture to being created niques are Methods models. aspects in climate of ETCs minimum or as a ETCs identify commonly ------., 2013; ., 2003; ., 2017). ., 2003).As et al et al et al ., 2013; Lakkis et al ., relative 2017), et al ., 2015; Massey, 2016; et al ., 2006). Massey’s (2012; et al et et al et ., 2008; Neu ., 2013; Michaelis ., 2015; Catto, 2016; Michaelis et al et al et al ., tracking 2017). Using algorithms ., (Raible height 2017), geopotential ., pressure 2008) and mean sea-level ., to often is reduced Vorticity 2013). ., 2019). ., 2015; Yang The trackingThe Hodges (1994; algorithm by Multiple climate variables can be used to to variables can be used Multipleclimate allows for the analysis of long-term trends trends of long-term the analysis for allows along with their of ETCs, and the lifecycle Hodges, 2002). this method com Although it does quick and simple statistics, putes of detail about ETC the level not provide such as the number and characteristics, determine used to are that of ETCs, intensity or impacts trends (Hoskinschanges in ETC and Hodges, 2002; Anderson Zappa involves however, method, Lagrangian The tracking and spatial of an the temporal knownindividual ETC, as objective fea trackingture (Hoskins and Hodges, 2002; Feser et al (Hoskins and Hodges, Neu 2002, 2005; et al the amount decrease to resolution a lower of noise (Hoskins and Hodges, 2002, 2005). Tracking used frame commonly two are There in cli tracks storm evaluating for works and Lagrangian. models: Eulerian mate uses a method commonly Eulerian The syn highlight 2–6-day to filter bandpass includes which activity, timescale optic (Blackmon, 1976; tracks storm Hoskins and 1995; 1999) uses relative vorticity at 850hPa vorticity 850hPa at 1995; 1999) uses relative and has fre of ETCs the identification for been used in feature-trackingquently stud ies (Bengtsson 2016) objective feature-tracking algorithm uses re-gridded iden minimum MSLP to these Using latitudes. higher at ETCs tify in identification approaches different two outputted the in variations to lead can this for One reason statistics. track storm the using MSLP represent results is that the of large-scale features low-frequency, vorticity whereas represents atmosphere, small-scale features the high-frequency, vorticity (Hodges, 1995; Zappa Chang, kinetic 2017), eddy (Wang energy et al et al (Hoskins(MSLP) Hodges, and Feser 2002; al et Chang, choice frequent most The 2017). local minima in MSLP or use either is to maxima in vorticity a single geopoten at (in the mid– level or pressure tial height and an ETC identify to troposphere) lower time and space through feature that track (Raible et al and Hodges, 2002; Anderson achieve to way standardised is no set there meth and some removal, this background can results step, such a involve ods do not vary another. one method to from including meridi position, an ETC’s identify onal winds (Booth ------., 2009). et al ., Feser 2013; et al ., inter-seasonal 2010). On ., 2014). et al et al ., 2013). It is agreed, however, that that ., 2013). however, It is agreed, et al ., Zappa 2003; ., 2015; Massey, ETCs allows This 2016). Before the identification and tracking the identification of Before The multitude of various dynamics that that dynamics various of multitude The towards slower-moving systems (Hoskins systems slower-moving towards et al as extrema from easily identified be more to bias any and removes systems larger-scale ETCs, many storm-tracking algorithms many apply ETCs, spa the large which remove filters, spatial tial scale or small-noise scale (Anderson et al ing the number of storm tracks crossing a crossing tracks ing the number of storm time (Ulbrich through region the number of ETCs is simply the number of the number of ETCs frequency is ETC in the data, ETCs identified and in a defined area, the number of ETCs count by density measured be can track is no universally agreed definition of what definition of what agreed is no universally is location its precise is or where an ETC (Neu Each tracking algorithm has a set of known when trying iden to overcome obstacles to observa and model the within ETCs tify there is that One such problem tional data. ETC tracking ETC Identification case study of two strong ETCs in the North ETCs strong case study of two Atlantic. literature where studies have compared a a compared have studies where literature data several using statistics ETC of range a by illustrated sets and methods follows, identified systems through time and the time and the through systems identified them, and the overcome to ways different of definitions of using different significance of the current A review extreme or intense. ally growing. The paper first describes the The ally growing. identification. multiple methods used for in tracking the obstacles then explore We three methods. More emphasis has been emphasis has been More methods. three the on NH tracking due to placed results however, of literature; availability greater continu is tracks on the SH storm research range of ETC statistics using several data several using statistics ETC of range two sets and methods and (4) compare using tracks North ETC transitional Atlantic tracking ETCs, (2) discuss the implications the implications (2) discuss tracking ETCs, definitions of extreme or of using different overview an of the current give (3) intense, a compared have studies where literature ous and diverse trackingous and diverse methods; there overviewan give (1) to aims paper this fore, and in identifying of the methods used can control storm track characteristics track pre storm can control in how challenge a significant us with sents their impactsand understand measure we in numer has resulted This our world. across ward shift of SH storm tracks during winter shift during winter tracks of SH storm ward (Lehmann North Atlantic and Pacific storm tracks move move tracks storm North and Pacific Atlantic returning before summer, in the poleward (Hoskins equatorward in the winter and Hodges, is a pole there 2019). Similarly, directing ETCs towards northern Europe northern towards Europe directingETCs (Woollings of the in the NH, the latitude timescales Hodges (1995), repeating the process of ing algorithm. They found a future basin-wide ference was found during summer with identifying and tracking an ETC using rela- reduction in the number of strong ETCs dur- additional discrepancies in the number tive vorticity on multiple vertical levels and ing winter. However, an increase in number and intensity of tracks in regions close to then stacking these results to create a 4D and strength over the British Isles and central significant orography. In 2009, Ulbrich et al. representation of the track. Europe was projected. In addition, Michaelis reviewed various methods of identification All these different approaches to tracking et al. (2017) investigated the impact of cli- and tracking using different reanalysis data- can influence the calculation of ETC char- mate change on the winter North Atlantic sets for both hemispheres. They also found acteristics and statistics (Feser et al., 2015). storm track and found an overall decrease that most disagreements were for summer It is important to note that each method in the number of strong ETCs in the North months, and there was a better agreement has its limitations, and there is no ‘correct’ Atlantic when defining strong as passing a for intense ETCs. The differences when way to solve these issues. As a result, it is minimum threshold in the sea-level pressure comparing reanalysis datasets were mostly not advised to apply an algorithm without field. Alternatively, in the SH, Chang (2017) related to the different spatial resolutions. The numerous approaches to tracking extratropical cyclones extratropical tracking to approaches numerous The knowing its limitations. found that a significant increase in the fre- The role of uncertainties due to the quency of future extreme ETCs was not tracking method was highlighted by Neu dependent on the definition used. et al. (2013), who assessed 15 different track- Defining ‘extreme’ ing algorithms as part of an experiment set Just as there are many different climate up by the international Intercomparison of variables used in identification, there are Comparing methods MId LAtitude STorm diagnostics (IMILAST). numerous methods of defining and classi- Differences due to datasets and The experiment was set up so that each fying what is an ‘extreme’, ‘strong’ or ‘intense’ tracking algorithm used the same dataset ETC (Catto, 2016; Chang, 2017). Approaches methods (ERA-Interim reanalysis) for the same period can involve defining extreme in terms of There are differing results in climatological (1989–2009), at the same spatial (1.5° × 1.5°) passing a physical threshold, and others storm-track structures and densities and and temporal resolution (6-hourly time- account for the physical damage caused in historical and future trends. These may steps). The largest differences between by an ETC, whereas some combine these result from differences in the data used or methods were for the number of ETC Weather – November 2020, Vol. 75, No. 11 No. 75, Vol. 2020, – November Weather (Garnier et al., 2018). Lambert (1996, pp 21, differences in the methodology of tracking tracks. There was a larger spread of results 320) defined an intense ETC as ‘the occur- ETCs. Uncertainties regarding the dataset in the NH and over continents than in rence of a grid point value of MSLP less were identified by Hodges et al. (2003), who the SH. Figure 2 shows the differences than or equal to 970mbar’. This threshold used several reanalysis datasets, together in the number of NH (30°–90°N) ETCs for was used to ensure the exclusion of most with Hodges’ (1999) tracking algorithm, December, January and February for each ETCs, spurious lows and any low pres- to compare the representation of his- of the methods used, with additional results sures caused by high terrain. Alternatively, torical storm tracks in both hemispheres. from Massey (2016). These methods differ Zappa et al. (2013) defined strong ETCs as Differences between the reanalyses were by more than 100%, with no well-defined exceeding the 90th percentile of maximum greater in the SH, in regions of growth or grouping of results based on climate vari- wind speed at 850hPa in the North Atlantic decay, and were generally larger for weaker ables. However, there was more agreement and European storm tracks. More recently, ETCs. Fewer observations in the SH generate in the number of winter ETCs identified in Chang (2017) applied different definitions a greater dependence on model results and both hemispheres. Winter ETCs tend to be of extreme based on the exceedance of consequently increase the uncertainty of more intense and easier to identify and two set thresholds using variables such as historic trends (Hodges et al., 2003; Ulbrich track, which is in agreement with Hoskins MSLP, 850hPa relative vorticity and winds. et al., 2009). Raible et al. (2008) compared and Hodges (2002, 2005) and Ulbrich Conversely, Grieger et al. (2018) defined NH ETC statistics between two reanalysis et al. (2009). Grieger et al. (2018) used the extreme as the top 500 most intense winter datasets for the period between 1961 and same approach as Neu et al. (2013) to fur- tracks when using minimum MSLP to meas- 1990. Although results for extreme ETCs ther understand the SH results. They found ure intensity. To help reduce discrepancies were in good agreement, the greatest dif- many similarities between methods, but like between assigned intensities, it is common to use MSLP (Feser et al., 2015). It is important to understand that differ- 25 000 ences may arise in trends when the definition of what represents an extreme ETC is s 20 000 e

not consistent. This is not only relevant for n o l

historic trends but also for future projections c y 15 000 c

as numerous studies use different definitions f o r

of ETC intensities (Ulbrich et al., 2009; Zappa e 10 000 b

et al., 2013; Michaelis et al., 2017). Research by m u

Ulbrich et al. (2009) showed that the results N 5000 of future hemispheric trends in extreme 0 ETCs depended on how they were defined. M02M03 M06M08 M09M10 M12M13 M14M15 M16M18 M20M21 M22M23 A decrease in the number of extreme ETCs Method averaged over the whole NH was found when extreme was defined as being in the 99th MSLP VORT Z850 MSLP +VORT Z850 percentile for the Laplacian of pressure, com- Figure 2. The number of Northern Hemisphere (30°–90°N), winter (December, January and pared to an increase when defined in terms February) cyclones between 1989 and 2009, using 16 different tracking methods. Adapted from of sea-level pressure. Zappa et al. (2013) Table 2 in Neu et al. (2013), with the addition of M23, representing the results from Massey’s (2016) used a multi-model approach to investigate tracking algorithm. Grey dashed line represents the mean, with shading representing the primary the North Atlantic ETC response to RCP4.5 variable used for identification – blue represents MSLP, red represents relative vorticity at 850hPa and RCP8.5 future climate scenarios using (VORT Z850), hatched represents MSLP and/or Laplacian of MSLP or vorticity (VORT), and green Hodges’ (1995; 1999) objective feature-track- represents geopotential height at 850hPa (Z850). (Source: Neu et al., 2013.) 338 The numerous approaches to tracking extratropical cyclones

Neu et al. (2013), differences included varia- used; rather, it was the variations in thresh- ETC. Despite the MCMS track beginning tions in ETC numbers and intensity, with a old settings that were more significant. earlier than Massey, it diverges for a small greater agreement in intense ETC statistics. section when compared to the other tracks. As previously discussed, MSLP and rela- As Ophelia hits the British Isles, the differ- tive vorticity are popular climate variables Comparison of three methods ence between tracks decreases; however, used in feature tracking. Differences in the on two North Atlantic transitional they begin to separate towards the end location and number of tracks could be of the tracks. All three tracks finish in dif- due to the choice of variable used (Raible ETCs ferent locations, with MCMS at a different et al., 2008). When using MSLP and 850hPa To illustrate the variations that can occur time-step. Interestingly, the three methods vorticity, Hoskins and Hodges (2005) found when using different methods, two North demonstrate a closer agreement for Oscar’s that the SH storm track was strongest dur- Atlantic transitional ETCs, Ophelia (2017) track (Figure 3b). Once more, the largest dif- ing winter. However, when using 250hPa and Oscar (2018), were tracked using three ferences are at the beginning and end of vorticity (upper troposphere), maximum separate tracking methods (Figure 3). First, the tracks, with the MCMS method identi- values occurred during summer. Vorticity the National Hurricane Center (NHC) best fying and tracking Oscar before the NHC. and MSLP results agreed that the strong- track was obtained from its hurricane data- Agreement between the tracks improves Weather – November 2020, Vol. 75, No. 11 est ETCs occur in the southern Atlantic and base HURDAT2 (Landsea and Franklin, 2013). towards the latter half of the storm track. Indian Ocean regions. In addition, Grieger The NHC best tracks are created by collating However, as highlighted by the arrow in et al. (2018) found that vorticity identified all the observational data available, such as Figure 3(b), there is a noticeable outlier in a greater number of tracks in the SH than satellite and aircraft measurements, to sub- the Massey track. There is a closer agree- MSLP. This is a result of vorticity being more jectively determine the location, intensity ment between Massey and NHC regarding capable of identifying and tracking small- and size of tropical cyclones and their tracks. the location of dissipation than with MCMS, scale features (Hoskins and Hodges, 2002; Second, the National Aeronautics and Space which extends the track northeastwards by Neu et al., 2013; Grieger et al., 2018). It may Administration (NASA) Modeling, Analysis another time-step. be assumed that regions are dominated by and Prediction Climatology of Midlatitude Large differences between tracks can small-scale systems when there are more Storminess (MCMS) tracking algorithm exist at the initial and final time-steps for ETCs identified by vorticity than MSLP. applied a closed contour method using MSLP a variety of reasons, one being that rela- There are many similarities between MSLP minima from ERA-Interim to locate and track tive vorticity is more capable of identify- and vorticity tracks in the NH, except for ETCs (Naud et al., 2012). Finally, the Massey ing a system at an earlier stage than MSLP regions such as the Mediterranean and at tracks were created by inputting 6-hourly (Hoskins and Hodges, 2002; Neu et al., 2013; the beginning and end of tracks (Hoskins MSLP data from ERA5 reanalysis (Hersbach Grieger et al., 2018). Rantanen et al. (2020) and Hodges, 2002). Pinto et al. (2016) et al., 2020) into the Massey (2016) storm- tracked Ophelia using Hodges’ TRACK algo- discovered that ETC clustering in the tracking algorithm. rithm (1994; 1995) with input data from the North Atlantic and Europe compared well All the methods indicate a comparatively Open Integrated Forecast System model between multiple methods. However, there good agreement between the locations and compared it to the NHC best track. was less agreement around the initial and of the two tracks. There are apparent dis- TRACK identified Ophelia earlier than both final positions of the storm tracks, with similarities, especially at the initial and final MCMS and Massey and had a similar dissipa- vorticity tracks being located further south stages of the tracks. The most evident differ- tion location as MCMS. This highlights that than MSLP tracks. Conversely, Hewson and ence in Figure 3(a) is that the NHC best track methods using vorticity and MSLP can pro- Neu (2015) stated that they could not group begins much earlier, identifying Ophelia as duce a comparable track, with less similarity their results based on the climate variable a hurricane before it transitioned into an shown at either end of the tracks.

Figure 3. The National Hurricane Center (NHC) best track (red circles); the Modeling, Analysis and Prediction (MAP) Climatology of Midlatitude Storminess (MCMS) track (blue triangles); and the Massey (2016) track (black crosses) for Storm Ophelia, 16 October 2017 (a), and Storm Oscar, 3 November 2018 (b). Each point is at 6-hourly intervals and the mean sea-level pressure from ERA5 reanalysis is plotted for the time-step indicated in each figure. 339 340

Weather – November 2020, Vol. 75, No. 11 The numerous approaches to tracking extratropical cyclones in itslocation, even between thetwo MSLP there was some disagreement (Figure 3(a)), peak intensity interms ofminimumMSLP it isinteresting that whenOpheliareached that are reasonably easierto track. Therefore, Consequently, theyrepresent stronger ETCs andthentransitioned tohurricanes ETCs. tropical transition, inthat theybeganas ments (LandseaandFranklin, 2013). intensity andavailability ofaircraft measure in theHURDAT2 dataset dependedonits overThe acyclone’s uncertainty position oftheinformationand quantity available. to itisrestricted thequality observations, statistics. While theNHCbesttrack uses ‘right’ answer whenanalysing storm track and cyclolysis. there Nevertheless, isno interms oflocationpicture ofcyclogenesis in methodscanpaint aslightly different shows how changes and tracking. Figure 3 approaches and thresholds in identification MSLP, they used different reanalyses, • • • • sions from thisstudy are asfollows: will changeinthefuture. The mainconclu how their numbers, intensities and impacts makes iteven more difficultto agree on ETC trends have been in the past, which may leadto alack of consensus onhow we propose that thisdiversity ofmethods one singleagreed-upon method. As aresult, lenging to fullyrepresent ofETCs alltypes in due to their complexity, itisextremelychal measure itsintensity andlifetime. However, approaches to identify andtrack anETC and ods are developed. There are now more ETCs meth isincreasing asnew tracking The oftracking diversity andcomplexity and conclusion Summary methods.

Both Ophelia and OscarcompletedBoth extra usingDespite both MCMS and Massey The largest differences meth intracking The useof different datasets hasshown Differences inthedefinitionof Despite many differences, there are of ETCs. theNH,eachmethod used In ods exist whenexamining thenumber improve results. ETC tracking accurate andrealistic data, helpingto reanalysis are products providing more in greater agreement. Newandupdated or decay. for Results extreme ETCs are weaker ETCs andinregions ofgrowth exist summer, in the SH,during for that thelargest differences inETC tracks different definitionsof cautious whencomparing results using ence at all. Therefore, onemustremain signs oftrends ornothave any influ tant role astheycanresult inopposite ETCs have beenshown to play animpor on their duration and distance travelled. ods that involve filtering tracks based some common features amongmeth extreme . extreme ------

• authors declare ofinterest. noconflict Research Council (NE/L002434/1). The studentship from the Natural Environmental GW4+Doctoral NERC Training Partnership bypublicly available. a issupported EW datalike to ERA5 thank ECMWFfor making code andwould hisstorm-tracking sharing this paper. The for authors thank Neil Massey ers for theirhelpfulcomments onimproving The authorsthanktheanonymous review Acknowledgements et al (Raible tinue to compare methods storm-tracking we agree that there isstillaneedto con method. use onlyonetracking Therefore, when examining trends inETC statistics that may produce significantly different results algorithm same dataset onanothertracking consider that usingadifferent dataset orthe physical systems. itiscrucialto Nevertheless, growing understanding of these complex nificant information that hashelpedinour References

Each study has contributed new and sig Bengtsson L Bengtsson Rev. background fieldremoval. methods ofstorm tracks to theform of analysis offeature-based 2003. Sensitivity Anderson D Booth JF Booth J. Atmos. Sci. Hemisphere. tial height oftheNorthern studyofthe500mbgeopoten spectral ML Blackmon 2276–2301. sify inawarmerclimate? N L Bengtsson J. Climate 2006. Storm tracks andclimate change. Brayshaw DJ 2539–2558. trast andorography. Atlantic Part I:land-seacon storm track. 2009. The basicingredients oftheNorth Brayshaw DJ J. Climate storm tracks inCMIP5models. surface Spatial patterns andintensity of the 2011. The basicingredients ofthenorth The greatest differences between tracks when using MSLP or vorticity istowardswhen usingMSLPorvorticity for summerandweaker statistics. winter andintense ETC statistics than There is, however, more agreement for 2). the climate variable used(Figure 100%, withnocleargrouping basedon in Neu tures. todue to track itsability small-scalefea can identify a higher number of ETCs the firstandfinaltime-steps. Vorticity ., 2013; Grieger Grieger ., 2013; . 2009. Will extratropical storms inten

131 et al : 565–573. et al

, ., 2008; Ulbrich Ulbrich ., 2008; Kwon YO 19 30

, , , : 3518–3543. : 4965–4981. , 33 , Hodges KI . (2013) differs by more than . (2013) Hodges KI Hodges KI Hoskins B Hoskins . 1976.Aclimatological Hoskins B Hoskins : 1607–1623. et al , J. Atmos. Sci. Ko S Ko ., 2018). , , , , , J. Climate

Hoskins BJ Hoskins Blackburn M Blackburn Roeckner E Roeckner Keenlyside et al Blackburn M Blackburn Mon. Weather et al . 2017. ., 2009; Neu ., 2009;

22 : . . : - . - . - - - - - fied by 14 tracking methods,fied by andtheirrole 14tracking et al J Grieger the U.S. andEurope. deck infavour ofharsh winter weather in tent. Loss the ofArcticseaice isstacking Greene C and Italy. case studiesinFrance, England, Portugal analysis ofstorm2018. Historical events: E Garnier Meteorol. Soc. review.and northwestern Europe-a Atlantic2015. Storminess over theNorth Feser F Hemisphere. extratropical intheSouthern cyclones cant increase inthenumberofextreme Chang EKM J. Climate high-resolutionHiGEM climate model. inawarmingclimatecyclones inthe Hemisphere extratropical Northern Catto JL pacific region. climate studies:testing inthesouthwest tively classifyextratropical for cyclones Catto JL Rev. Geophys. sification anditsuseinclimate studies. Catto JL temperatures. Atlantic Part II:seasurface storm track. the UnitSphere. Hodges KI 122 meteorological data. analysis anditsapplication totracking Hodges KI Meteorol.Dyn. Oceanogr. storms andtheIMILASTproject. Hewson TD Meteorol. Soc. 2020. The globalreanalysis. ERA5 Hersbach H Lett. with extratropical cyclones? Hemisphere precipitation is associated KI Hawcroft MK Tellus Meteorol. ADyn. Oceanogr. for into moisture thecontinent. transports Hodges KI 1362–1373. feature tracking. Hodges KI 3458–3465. oscillation. In oscillation. In Atlantic oftheNorth 2003. An overview Hurrell JW 32 tracks. Part II:regional detail. Hemisphere storm ofNorthern cycle BJ Hoskins tracks. Hemisphere onSouthern storm spective BJ Hoskins 1041–1061. winter storm tracks. Hemisphere ontheNorthern spectives BJ Hoskins Mon. Weather Rev. storm tracks andtropical easterly waves. datasets feature usingobjective tracking: 2003. Acomparison ofrecent reanalysis

: 1761–1775. et al : 2573–2586.

. 2018. Subantarctic cyclones identi . 2018.Subantarctic cyclones 39 J. Climate . 2012. How much Northern . 2012.How muchNorthern : 1–7. , , . 2018.Anewmethodto objec clas . 2016.Extratropical cyclone Barcikowska M

, Shaffrey LC Coast. Eng. . 2012. The winters ofourdiscon , 24 Leckebusch GC , . 1999.Adaptive constraints for . 1995.Feature Tracking on . 1994.Ageneral methodfor Ciavola P , , , , Hoskins BJ Hoskins , Kushnir Y . 2017.Projected signifi , Hodges KI Hodges KI Hodges KI Neu U : 5336–5352.

J. Climate Bell B Bell

The North Atlantic Oscillation: Atlantic Oscillation: The North , 146 54 J. Atmos. Sci. 141 J. Climate Shaffrey LC

Mon. Weather Rev. Mon. Weather Rev. : 486–520. 18

: 1999–2049. : 350–382. 131 , . 2015.Cyclones, wind : 4108–4129.

J. Atmos. Sci. Berrisford P Sci. Am. Sci. , Mon. Weather Rev. 134 , Spencer T ,

Hodges KI : 2012–2037. Ottersen G Ottersen , . 2019. The annual . 2005.Anewper . 2002.Newper 30

Boyle J , 31 : 10–23. 67 Krueger O : 4915–4935.

, , 68 : 4683–4704. Raible CCRaible Hodges : 27128

Geophys. Res. J. Climate 307 : 1784–1805.

Tellus A

et al et al

70 : 50–55. Q. J.R. . 2011. 59 et al

- et al 123 Q. J.R. : 1–18. 127 :

et al . .

- - : : . - - - . - The numerous approaches to tracking extratropical cyclones Weather – November 2020, Vol. 75, No. 11 341

2 - - - : . . 28 . :

et al

. 2018. . 2018. : 2786–2794. Hoskins B J. Climate , 45 Q. J. R. Meteorol. J. R. Meteorol. Q. J. Climate Hodges KI

, Gudgel RG , Chang EKM Chang , Hannachi A , Shaffrey LC Shaffrey : 856–868. , Kim HM Vecchi GA Vecchi , , Geophys. Res. Lett. Res. Geophys. 136

5846–5862. Hemisphere winter storm tracks by the by tracks storm winter Hemisphere QBO. T Woollings of the North Variability Atlantic 2010. eddy-driven jet stream. Soc. X Yang 2015. Seasonal predictability of extratrop high-resolution in GFDL’s tracks ical storm prediction model. climate 3592–3611. Zappa G of 2013. A multimodel assessment projections of Northfuture and Atlantic cyclonesin the extratropical European models. CMIP5 climate Wang J Wang of Northern modulation Interannual Measuring road traffic emissions of NO Measuring traffic road at high spatial and temporal resolutions is resolutions and temporal high spatial at In this challenging. and logistically costly a cheaper and univer develop we study, road infer sally applicable methodology to transport of an the resolution at emissions amount utilise the vast We individual road. use the widespread by generated of data understand better of mapping products to the total While on city roads. flows traffic link is on a busy road number of vehicles net emissions data from road transport road is from data emissions net the 1-km at from available resolution freely Emissions Inventory Atmospheric National this top-down approach (NAEI). However, loses the granularity of particular and roads dif emission hotspots at junctions are that times of the day. ferent of the Creative Commons Attribution License, License, Attribution Commons of the Creative distribution and reproduc which permits use, original the medium, provided tion in any cited. is properly work doi: 10.1002/wea.3861 Correspondence to: Erin Walker to: Correspondence [email protected] published by Weather Authors. © 2020 The & SonsJohn behalf on Wiley of the Royal Ltd SocietyMeteorological article is an openThis access under the terms - - - - . - : lev : - - ) and and ) 2 - 2 . -

et al 25

. 136

et al

30 Theor. Theor.

), with 64% ), with 64% . 2017. 2.5 Bull. Am. Bull. Schwierz Schwierz Pinto JG Pinto , , J. Climate Sinclair VA , ., 2019). The pri The ., 2019). : 1629–1642. Van Den Heever Heever Den Van J. Climate , Bellenbaum N , 49 Economou T Economou in roadside locations locations in roadside

, 2 et al et Tellus A Dyn. Meteorol. A Dyn. Meteorol. Tellus Chang EKM Chang Mon. Weather Rev. Mon. Weather : 117–131. Tellus A Dyn. Meteorol. A Dyn. Meteorol. Tellus , : 529–547. 96

94 Räisänen J : 1–26. : 1–15.

. 1992. Midwinter. 1992. suppres , Della-Marta PM 72 68 Posselt DJ Posselt Leckebusch GC Leckebusch , , ,

Ulbrich S , Kim HM , Akperov MG Akperov J. Atmos. Sci. J. Atmos. , . 2008. Northern Hemisphere . 2020. The extratropical transition of transition extratropical The . 2020. . 2012. Observational analysis of . 2012. Observational of analysis et al

Wang J Wang Changes in Northern win Hemisphere the background under tracks storm ter of arctic amplification. 3705–3724. et al Hurricane (2017) as diagnosed Ophelia and omega equation with a generalized vorticity equation. Oceanogr. Ulbrich U cyclones2009. Extra-tropical in the pres a review. climate: and future ent Climatol. Appl. 2016. Robustness of serial of clustering of cyclones the choice to extratropical tracking method. Oceanogr. Raible CC C cyclones: of a comparison extratropical detection and tracking methods and dif reanalyses. ferent 880–897. Rantanen M 2013. IMILAST: a community effort to 2013. IMILAST: cyclone detec extratropical intercompare tion and tracking algorithms. Soc. Meteorol. JG Pinto Nakamura H Nakamura activity in the wave sion of baroclinic Pacific. Naud CM SC midlatitude in cloud and precipitation cyclones: Northern Southern versus warm fronts. Hemisphere 5135–5151. U Neu The sources and profile of pollutants var of pollutants and profile sources The els (Achakulwisut els (Achakulwisut mary of NO source Yearly transport. vehicle from originates UK alone (Royal College of Physicians, 2016). of Physicians, College UK alone (Royal popula than 50% of the world’s more As it is urbanin environments, resides now tion in which areas small spatial these relatively likely pollution episodes are the most acute impact the largest on the greatest have to number of people. In the UK, the world. throughout ies greatly health drivingthe key pollutants adverse (NO dioxide nitrogen are outcomes fine particulate (PM matter of new paediatric asthma cases in urban NO elevated to attributed centres nomic burden of £20 billion per year in the of £20 billion per year nomic burden - - - -

. - - - . - -

Mon. et al

et al : 1–23.

: 1–8. 9

145

: 433–440. 8 Lackmann . 2013. Atlantic . 2013. Atlantic

: 6905–6925. , Frieler K Frieler , . 1975. The effects The . 1975. , Yuchechen A Yuchechen , 30 Hardiman SC Hardiman

, : 42–51. : 36–44. J. Geophys. Res. Atmos. Res. J. Geophys. Using routing apps to model apps to routing Using : 3576–3592. 44 93 : 3–15.

., 2017), while contributing ., 2017), while contributing Willison J Willison Environ. Res. Lett. Res. Environ. Franklin JL Franklin Nat. Geosci. 141 , 32 , Coumou D Coumou J. Climate

. 1996. Intense extratropical . 1996. Intense extratropical Wetherald R Wetherald , . 2016. Feature tracking in . 2016. Feature Canziani P Canziani Scaife AA real-time road traffic emissions traffic road real-time , , et al . 2012. Feature tracking on the . 2012. Feature , Q. J. R. Meteorol. Soc. Meteorol. J. R. Q. . 2017. Changes in winter North. 2017. Changes in winter , Hurrell JW, Kushnir Y, Ottersen G, Y, Kushnir JW, , Hurrell . 2019. A 4D feature-tracking algo et al . (eds). American Geophysical Union: Union: . (eds). American Geophysical : 21319–21325.

000 premature deaths and an eco an and deaths premature 000

GM cyclones in high-res extratropical Atlantic pseudo-globalolution regional warming simulations. Massey NR Massey data. climate regional high-resolution Geosci. Comput. Michaelis AC Massey N Massey triangular mesh. equal area hierarchical Geosci. Comput. Manabe S on concentrations of doubling the CO2 model. circulation of a general the climate Sci. J. Atmos. Lehmann J Lehmann changes in extratropical 2014. Future under cli and baroclinicity tracks storm change. mate Landsea CW hurricane uncertainty database and pre format. a new database of sentation Rev. Weather Lambert SJ Northern cyclone winter Hemisphere 1899-1991. events: 101 Lakkis SG et al rithm: a multidimensional view of cyclone systems. Kidston J Kidston on tropo influence 2015. Stratospheric and sur tracks storm spheric jet streams, weather. face Climatic Significance and Environmental and Environmental Significance Climatic Impact et al DC, pp 1–35. Washington, Birmingham, Edgbaston, UK Birmingham, Edgbaston, Environmental Sciences, University of Sciences, Environmental to 40 to the above; poor air quality disproportionthe above; ally impacts minorities and those on a low (Di income Introduction pressing face world the around Countries political and eco environmental, social, of all quality transcends Air issues. nomic Cai and William Bloss William and Cai Earth and School of Geography, Helen Pearce Pearce Helen Zhaoya Gong, Xiaoming