1. Introduction
Midlatitude synoptic-scale cyclones and anticyclones are known to play a central role in the atmospheric general circulation through their transport of heat, moisture and momentum (see Grotjahn 1993, 342–374; Chang et al. 2002; Shaw et al. 2016). Further, the frequency of occurrence and intensity of mobile synoptic-scale cyclones is an important component of the weather and climate in the midlatitudes. Global economic damage from extratropical cyclones is on the order of $7 billion (U.S. dollars) annually and may increase by an annual $2.8 billion with climate change (Narita et al. 2010). These estimates are now decades old and would be considerably higher in today’s dollars.
As a consequence, there is interest in characterizing these cyclone influences with documentation of cyclone histories—their intensities and tracks throughout their lifetimes. There are two general approaches for obtaining this information. One common technique is to consider synoptic-scale storm tracks as a statistical feature of the climate (e.g., Swanson 2007), and define them on gridded analyses through the variance of bandpass filtered geopotential height fields (e.g., Blackmon 1976; Wallace et al. 1988), where the time-filter is applied for synoptic time scales. Raible et al. (2008) point out, however, that a substantial portion of this resulting variability is related to large-scale waves (∼8000 km) rather than synoptic-scale (∼3000 km) cyclones (see also Doblas-Reyes and Déqué 1998). Further, both troughs and ridges are considered together in this method, which is a difficulty because these features are often asymmetric and because clouds and precipitation typically occur downshear of cyclones and upshear of anticyclones. In addition, there is no means for examining the life cycles of individual cyclones. For these reasons, the second general approach, that of identifying and tracking individual cyclonic systems, is considered in this study.
The most direct method for acquiring cyclone track information is through the labor-intensive process of manual tracking (e.g., Petterssen 1956; Klein 1957; Reitan 1974; Colucci 1976; Zishka and Smith 1980; Sanders and Gyakum 1980; Whittaker and Horn 1981; Roebber 1984; Gyakum et al. 1989). One significant advantage of this method is that the process of sifting through these data allows an investigator to notice peculiarities, such as a secondary development, which might otherwise be missed. Furthermore, while the manual process is intrinsically subjective, this may be a benefit where the analyst is synoptically experienced and can make sometimes complex tracking decisions informed by that experience. Such decisions are typically documented in the form of cyclone tracks drawn by an analyst directly on the maps and are featured in individual case studies.
Neu et al. (2013) report on Intercomparison of Mid-Latitude Storm diagnostics (Project IMILAST; see http://www.proclim.ch/imilast/index.html), which examined 15 different automated methods of detection and tracking of extratropical cyclones using a common analysis dataset (the 1989–2009 ERA-Interim reanalysis). These authors noted that, while there is no accepted single “truth” concerning cyclone tracks, careful manual tracking might be considered as optimal.
The primary and substantial advantages of automation, in our view, are twofold. First, automation provides the ability to readily increase sample size. Roebber (1984) tracked all cyclones in a large portion of the Northern Hemisphere but was only able to do so for 1 year of data. Gyakum et al. (1989) followed a similar methodology, but by employing multiple analysts, was able to enlarge their study considerably. Today, using automated approaches, either study could be repeated relatively easily.
A second advantage of automated methods is the guarantee of self-consistency. Different analysts, when confronted with the same data, may produce unique results, especially when the task is complex and not well-defined. This has been shown to occur in frontal analysis (e.g., Uccellini et al. 1992; Sanders and Doswell 1995), although to our knowledge no attempts have been made to determine the degree of inter-analyst reliability for manual cyclone tracking. Gyakum et al. (1989) attempted to enhance manual tracking consistency by noting and cross-checking the cyclone tracks on each 6-hourly map, but nonetheless it should be expected that manual approaches will not attain the self-consistency of an automated method.
We note, however, that self-consistency is not identical to objectivity, because any approach, automated or otherwise, requires a number of choices to be made, beginning with what will define a cyclone [see Schemm et al. (2018) for a discussion of this issue]. Next, the questions arise as to which variable to analyze (e.g., mean sea level pressure as per manual analyses or vorticity, as is often used in automated methods) and at what level (e.g., mean sea level, 850 hPa).
For example, the advantage of vorticity as the variable of choice is that, because it is proportional to the Laplacian of the height or pressure field, it may more readily detect early stages of cyclone development. A disadvantage that arises in this instance, however, is the resulting “false positives” that reflect small-scale variations in a height or pressure field that are not early-stage cyclones as understood by synoptic analysts. Controlling this effect will typically necessitate some form of smoothing or filtering of the vorticity field, which results in the loss of small-scale detail and thus potentially in the late detection of cyclones. Rudeva et al. (2014), as part of the IMILAST project, considered 13 different automated tracking methods, and found that late detection of cyclones could lead to reductions in cyclone numbers of up to 40% in the Northern Hemispheric oceans, and that about one-third of their cyclone sample attained rapid intensification during those critical early stages. Simmonds and Rudeva (2014) performed a similar analysis for extreme cyclones in the Arctic basin and found substantial differences in intensification rates.
Both manual and automated tracking approaches first identify cyclones at a given time and then use some form of “map” continuity to link the set of cyclones from one time to the next. Automated identification uses rules that pinpoint relative minima in pressure or geopotential height (e.g., Blender et al. 1997; Blender and Schubert 2000; Gulev et al. 2001; Wernli and Schwierz 2006), maxima (Northern Hemisphere) or minima (Southern Hemisphere) in measures of vorticity (e.g., Murray and Simmonds 1991; Hodges 1994; Sinclair 1994, 1997) or some combination of the two (e.g., König et al. 1993; Hewson and Titley 2010). As noted above, manual cyclone tracking involves a degree of subjectivity where decisions are informed based upon synoptic experience concerning cyclone behavior and dynamics. Translating this into an objective process is not straightforward and as a result, a variety of methods have been developed to accomplish this task [see Ulbrich et al. (2009); Neu et al. (2013); Ullrich and Zarzycki (2017) for overviews of automated methods]. It is important to recognize, as noted above, that while automated methods are self-consistent in the sense that the same input data with the same settings will provide the same tracking results, specification of algorithm parameters must be made either at design or upon implementation, and it is not obvious which automated scheme is superior (e.g., Raible et al. 2008; Ulbrich et al. 2009; Neu et al. 2013; Ullrich and Zarzycki 2017).
Whether using manual or automated methods, the integrity of the input analysis is also of concern. Over maritime regions, the paucity of observations can lead to substantial differences in sea level pressure analyses, even between two objective analysis systems relying on similar data (Swanson and Roebber 2008). Raible et al. (2008) compared the variability in results obtained from three automated cyclone tracking methods using two reanalysis datasets (NCEP–NCAR and ERA-40). Reanalysis differences resulted in a systematically higher frequency of cyclone detection using ERA-40. Sanders (1990) compared manual and automated operational sea level pressure analyses with two sets of research surface analyses that used the intensive offshore observations collected during the Experiment on Rapidly Intensifying Cyclones over the Atlantic (ERICA; Hadlock and Kreitzberg 1988) as “truth.” This comparison revealed that the automated analyses tend to differ from high-quality manual sea level pressure analyses, both in terms of the position and the intensity of cyclonic systems, with the automated analyses showing weaker cyclones.
Model resolution used in automated analyses has increased significantly since 1990. Whether this resolution enhancement manifests itself in terms of improved cyclone tracking is unclear, although recent studies (Hodges et al. 2011; Rohrer et al. 2020; Crawford et al. 2021) find the biggest discrepancies with lower analysis resolutions. Raible et al. (2008) also found that using different tracking methods for a specific reanalysis (ERA-40) resulted in differences in cyclone track length. Note that all of the manual storm track studies cited above employed manual sea level pressure analyses rather than automated analyses constructed from model initialization procedures. Over a period of several decades, the operational sea level pressure analyses have evolved with varying amounts of data. Compare, for example, the manual sea level pressure analyses of the 1970s and 1980s with current ocean analyses (e.g., https://ocean.weather.gov/Atl_tab.php). This further motivates our special focus on the ERICA surface analyses with its exceptional level of offshore data (see section 2).
Together, these findings suggest that statistics concerning the initiation and dissipation of cyclones (including redevelopments such as often occur in association with western ocean warm currents offshore of the eastern United States and Japan; e.g., Roebber 1984) will be affected by the choice of tracking method. Indeed, in the comprehensive comparison of automated methods conducted by Neu et al. (2013), two key features emerged: the largest disparities existed between the total numbers of cyclones and the detection of weak cyclones, with consistency between methods being the greatest for strong cyclones rather than for weak ones. This is problematic, because as discussed below, at the onset of rapid cyclogenesis, such storms may exist only as frontal waves, and thus this key stage of development can easily be missed in automated climatologies. In particular, frontal waves are not fully represented in the model underlying the reanalysis data, a limitation that may be addressed in the next generation of convection resolving models. In the interim, frontal identification algorithms have been developed in an effort to address this problem (e.g., Simmonds et al. 2012).
Led in part by Sanders and Gyakum (1980), much observational and model-based research has been dedicated to understanding the dynamics and predictability of rapidly intensifying cyclones [see Uccellini (1990), Bosart (1999) and references therein for a review]. This research has led to the understanding that such cyclones owe their intensity in large part to the efficient interaction between adiabatic processes and the diabatic process of latent heat release (Uccellini 1990; Bosart 1999; Roebber and Schumann 2011), usually over a relatively short time interval (e.g., Roebber 1984). Rapid cyclone development appears to be sensitive to the upstream development of midtropospheric jet/fronts (Lackmann et al. 1997) and attendant circulations (Xiao et al. 2002), with initial condition errors connected to the upstream feature leading to rapid error growth at the early stages (Simmons and Hoskins 1979; Orlanski and Sheldon 1993; Zhu and Thorpe 2006). The early stages of the cyclone life cycle, when oceanic fluxes of heat and moisture often have the largest impacts on cyclone development (Bosart 1999), are closely linked to subsequent rapid development (see Bosart 1975, 1981; Uccellini et al. 1987; Gyakum 1991; Gyakum et al. 1992; Bullock and Gyakum 1993; Gyakum and Danielson 2000; Roebber and Schumann 2011; Schultz et al. 2019). Thus, proper detection and tracking of cyclones throughout their life history, but most particularly at their early stages, are important to developing a more complete understanding of their dynamics.
We emphasize that the present study is not intended to establish superiority of manual or automated methods. Rather, given that there will be differences in the data obtained from any two approaches, including two automated approaches (e.g., Ulbrich et al. 2009; Neu et al. 2013; Rudeva et al. 2014; Ullrich and Zarzycki 2017), and understanding that automated methods are a logistical necessity for tracking large quantities of cyclones, we seek to document the results obtained from one widely used automated method when applied to a set of well-known and carefully studied intense cyclogenesis events documented in the refereed literature. Because the impact from such storms is substantial, we believe that this is the most relevant as well as the most feasible approach toward this documentation, a step that we believe is essential as cyclone tracks and intensities obtained from automated methods are widely used for a number of applications (e.g., Trigo 2006; Raible et al. 2008; Ulbrich et al. 2009; Narita et al. 2010; Chang et al. 2012). Further, we test the sensitivity of these results to analysis resolution by conducting these comparisons on several reanalyses spanning horizontal grid spacings from 0.5° to 2.5° (see section 2), and as a final check, we consider results from the most recent reanalysis, the fifth generation ECMWF reanalysis (ERA5; Hersbach et al. 2020), which is available at 0.25° grid spacing.
Our approach here is to assess the representation of these five well-studied cyclones using standard settings from the automated method, such as those used in global storm track studies that provide physically realistic global climatologies of cyclones. Given the small number of cyclones examined here, one could tune the settings of the algorithm to properly capture the life cycles of these particular storms, but it would be at the expense of biasing the global climatology. Our intention here is to highlight the limitations of the standard settings of automated tracking algorithms, which are often applied to climate model simulations to characterize how cyclones across the globe might change over the twenty-first century (Chang et al. 2012; Priestley et al. 2020).
In this paper, our main focus is on tracking obtained from the sea level pressure field to provide direct comparison with manual tracking. However, as a final check, we examine tracking using the lower-tropospheric relative vorticity field from the most recent reanalysis with the highest spatial resolution (ERA5).
The most comprehensive investigation of this kind would be to perform a comparative study using a large variety of automated methods, similar to Neu et al. (2013), but where manual tracking of a set of well-known cases would form the baseline for that comparison. Such an effort, however, is a major undertaking and would require international cooperation among various groups with automated tracking algorithms. Here, as a first step in a less comprehensive but still useful approach, we have selected the Hodges feature tracking algorithm (Hodges 1994, 1995, 1999; hereinafter TRACK), because it is widely used, including in well-cited papers on the representation of storm tracks in the CMIP5 and CMIP6 global climate models (e.g., Chang et al. 2012; Priestley et al. 2020). While this makes our results less generalizable, we believe that performing this analysis with TRACK illustrates the problems we wish to highlight. We do this not to single out TRACK (which is a highly skillful algorithm that has been applied successfully to numerous research applications), but rather to discuss issues that we believe are inherent to automated methods in general. Automated methods are composed of three parts: the analysis, cyclone detection, and cyclone tracking. Hereinafter, when we refer to TRACK, this should be understood to be referring to cyclone detection and tracking, whereas we deal with the underlying analysis separately.
The organization of this paper is as follows. Section 2 will describe the input analysis datasets and configuration details for TRACK. Section 3 provides detailed comparisons of the set of cyclone events, including short summaries of the selected cyclones and results from experiments designed to elucidate the role of input data resolution on TRACK performance. Section 4 summarizes the results and provides suggestions for ways in which automated methods might be improved.
2. Data and method
Five well-documented Gulf of Mexico/North Atlantic cyclone histories are examined (Fig. 1): 18–20 February 1979 (The Presidents’ Day Storm; Bosart 1981; Bosart and Lin 1984; Uccellini et al. 1984, 1985, 1987; Whitaker et al. 1988); 12–15 December 1988 (ERICA IOP 2; Roebber et al. 1994; Lackmann et al. 1997, 1999); 4–5 January 1989 (ERICA IOP 4; Wakimoto et al. 1992; Pauley and Bramer 1992; Chang et al. 1993, 1996; Chang and Holt 1994; Manobianco et al. 1994; Reed et al. 1994; Rausch and Smith 1996; Cohen and Kreitzberg 1997; Liu et al. 1997; Alexander et al. 1998; Zou et al. 1998; Xiao et al. 2000; Parsons and Smith 2004); 19–20 January 1989 (ERICA IOP 5; Reed et al. 1993a,b; Blier and Wakimoto 1995; Wakimoto et al. 1995; Kuo et al. 1996); and 12–14 March 1993 (Superstorm 1993; Gilhousen 1994; Kocin et al. 1995; Caplan 1995; Huo et al. 1995; Uccellini et al. 1995; Bosart et al. 1996; Dickinson et al. 1997; Schultz et al. 1997; Huo et al. 1998; Desjardins et al. 1998; Alexander et al. 1999; Schultz and Steenburgh 1999).
Manual cyclone tracks for the five cases during the period of maximum 24-h cyclone central pressure fall: Presidents’ Day Storm (P storm), ERICA IOP 2 (IOP 2), ERICA IOP 4 (IOP 4), ERICA IOP 5 (IOP 5), and Superstorm 1993 (SS93). The “L” indicates the position of the cyclone central pressure minimum at 6-hourly intervals. Dates are indicated by DDHH, where DD is day and HH is hour (UTC).
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
Analyses for the ERICA storms benefit from the considerably augmented surface data collected during that field program [see Sanders (1990) and figures therein; the manual central pressures and tracking in the present study are derived from the set of analyses prepared by Sanders for the three ERICA cases; the full set of ERICA analyses are available from the lead author upon request]. The five surface cyclones all are associated with a maritime environment, characterized by strong near-surface baroclinicity, and subsequent interactions with a mobile upper-level cyclonic vorticity maximum.
Gridded reanalyses used as input to TRACK are the National Centers for Environmental Prediction–Department of Energy Reanalysis 2 (NCEP-DOE; Kanamitsu et al. 2002) at 2.5° horizontal grid spacing; the European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim; Dee et al. 2011) at 0.75° horizontal grid spacing; the NCEP Climate Forecast System Reanalysis (CFSR and CFSv2; Saha et al. 2010, 2014) at 0.5° horizontal grid spacing; and the ERA5 at 0.25° horizontal resolution (Table 1).
Reanalysis datasets used in this study. Shown are details on horizontal and vertical resolution, temporal interval, and the generation of the data assimilation system. Note that, even though it has hourly analysis, we use ERA5 at 6-h intervals to be input into TRACK so as to be consistent with the other reanalyses and the temporal interval of climate model output.
Details concerning TRACK are provided in Hodges (1994, 1995, 1999). Prior to the tracking, we remove the seasonal cycle from the sea level pressure field. As discussed in Penny et al. (2010), removing the seasonal cycle helps to isolate transient weather systems and avoids the aliasing of time-average features into the results. TRACK minimizes a cost function to produce a set of cyclone tracks for the region from 25° to 90°N, using a minimum intensity threshold of 4 hPa (relative to the background climatology, i.e., a 4-hPa anomaly from the seasonally varying background climatology specific to geographic location). Constraints are applied regarding maximum cyclone displacement between time steps and track smoothness. The set of tracks produced is filtered to remove cyclones with lifetimes of less than 48 h and/or track length less than 1000 km. We use TRACK to follow minima in the 6-hourly sea level pressure field over the domain for the period 1979–2017.
We apply TRACK to the sea level pressure field for each reanalysis. First, as is standard in previous studies (e.g., Hoskins and Hodges 2002), we spatially filter the sea level pressure field to wavenumbers 5–42 prior to the tracking to isolate synoptic-scale disturbances. We refer to this as the “smoothed” method (SMTH). Unfortunately, as will be shown below, smoothing reduces the ability of a high-resolution reanalysis to capture the full intensity of the often extraordinarily tight inner cores of deep maritime cyclones. Some studies (e.g., Jung et al. 2012) have used higher truncation for sea level pressure because this field is smoother than vorticity, which could have the potential to better capture full cyclone intensity. In our second approach, we do not spatially filter the sea level pressure field prior to the tracking. While this method would not be appropriate for establishing a climatology of synoptic-scale cyclones (as it identifies smaller-scale point vortices, among other features), it may be more appropriate to isolate the strength of individual extratropical cyclones. We refer to this as the “unsmoothed” method (NOSMTH).
For ERA5, we also show cyclone tracks based on the lower-tropospheric relative vorticity field. The tracking method is identical to the SMTH method for sea level pressure, except that the input field is the relative vorticity at the vertical model level closest to the 850-hPa pressure level (with the detection threshold of 1.0 × 10−5 s−1). Tracking vorticity on the model level is done to avoid biases that occur when tracking cyclones on the 850-hPa pressure level near high topography [see Grise et al. (2013) for further discussion of this issue].
We note that TRACK is often used with vorticity rather sea level pressure, and several studies suggest that this approach results in more frequent cyclone detections (e.g., Hoskins and Hodges 2002; Hodges et al. 2011; Vessey et al. 2020; Walker et al. 2020). However, in this study we are interested in cyclone track and intensity and this latter attribute, if it is to reference surface cyclone central pressure, must invoke sea level pressure in some way rather than vorticity. TRACK provides the location and magnitude of the extreme value (maxima or minima being followed) of the field being tracked, so it can only provide intensity based upon the field tracked itself. Thus, if one is using 850-hPa vorticity to track features, the algorithm will return the location and local maximum value of the 850-hPa vorticity feature. The user must apply some additional method to determine the intensity when using vorticity. For example, one could take the latitude/longitude coordinates of the cyclone identified by TRACK when using 850-hPa vorticity and then manually obtain the SLP value from the raw reanalysis data at that location. However, such a method does not ensure consistency with the location of the sea level pressure feature. Here, by tracking using sea level pressure, we employ a method that can readily be applied to large samples such as those obtained from a reanalysis dataset or a climate model simulation and ensures that the cyclone intensity is matched directly to the cyclone position.
3. Results
To understand the impact of the above setting choices, consider the results obtained for a representative day (1200 UTC 4 January 2014; Figs. 2 and 3) when TRACK is applied to the ERA-Interim reanalysis. It is readily apparent that both the smoothed and unsmoothed methods capture well-developed synoptic-scale cyclones, whereas the unsmoothed method also detects subsynoptic-scale sea level pressure features. This is particularly problematic near topography (Fig. 3, which shows the same time as in Fig. 2, but with a close-up view of Asia). This difference can be quantified over the period of the reanalysis (Fig. 4). It is apparent that the smoothed method is necessary to obtain relevant statistics for synoptic scale cyclones using these settings.
Smoothed (red open circles) and unsmoothed (red crosses) cyclone detections for 1200 UTC 4 Jan 2014 in the Northern Hemisphere as applied to the ERA-Interim reanalysis. Sea level pressure (no smoothing; no climatology removed) is shown by the shading. The circles and crosses are shown only for the cyclones that met the 48-h-duration and 1000-km-displacement criteria.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 2, but for a zoomed perspective of higher terrain areas across Asia.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
Smoothed and unsmoothed cyclone detections from the ERA-Interim reanalysis, for the period 1979–2017 (January, February, March, November, and December months only). Cyclones per month indicates the number of unique tracks (i.e., each cyclone is only counted once, even if it spends more than one 6-h period in a grid box). This count reflects the average number of cyclones whose centers pass through a given 2.5° latitude–longitude grid box in a given month during the indicated periods.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
We note also that differences in reanalyses can lead to differing results, and that these differences are not uniquely associated with increasing horizontal grid spacing (Fig. 5). When the smoothed method is applied to input data from three reanalyses, a similar number of wintertime cyclones over eastern North America and the western Atlantic Ocean is detected on average from all three reanalyses (Fig. 5, blue bars; see also Hodges et al. 2011). But, when the unsmoothed method is applied, the most wintertime cyclones are detected on average in the ERA-Interim reanalysis, even though it has a slightly coarser grid spacing than CFSR (Fig. 5, yellow bars). Note that we do not include the ERA5 reanalysis in this comparison, as running TRACK on the unsmoothed high-resolution (0.25° resolution) sea level pressure field for ERA5 was not computationally feasible.
Smoothed and unsmoothed cyclone detections over North America for three reanalyses used in this study: NCEP, ERA-Interim (ERAI), and CFSR. The time period is 1979–2017 (January, February, March, November, and December months only). North America is defined as 25°–65°N, 90°–45°W.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
a. Presidents’ Day cyclone
The Presidents’ Day cyclone of February 1979 is noteworthy for its initiation and rapid intensification along a coastal front offshore of North and South Carolina (Bosart 1981). Bosart’s analysis shows a closed cyclonic circulation with a central pressure of 1018 hPa at 1800 UTC 18 February 1979. At 0600 UTC 19 February, a synoptic-scale closed circulation was established with a central pressure of 1012 hPa. Rapid intensification continued through 1800 UTC 19 February, when hurricane-force surface winds flanked the cyclone center and the cyclone central pressure was estimated by Bosart (1981) to be 990 hPa. Thus, the maximum deepening rate for this storm is 1.64 Bergerons.
b. ERICA IOP 2
The surface cyclone was first analyzed at 1011 hPa offshore of the central Florida coast at 1800 UTC 12 December 1988, then propagated north-northeastward with a slow deepening to 1003 hPa at 1200 UTC 13 December and positioned east of extreme northern Florida. The cyclone deepened rapidly to a central pressure of 964 hPa by 1200 UTC 14 December, at a rate of 2.66 Bergerons.
c. ERICA IOP 4
The incipient cyclone was analyzed at a 996-hPa central pressure at 0000 UTC 4 January 1989. Its development occurred immediately following its genesis, with a steady fall of central pressure to 936 hPa by 0000 UTC 5 January, and located southeast of Sable Island, Nova Scotia. This rapid intensification (an extraordinary 3.60 Bergerons) was accompanied by an areal expansion of its cyclonic circulation to a region extending from northern Newfoundland southward to the Bahamas.
d. ERICA IOP 5
The initiation of a closed cyclonic circulation occurred at 0000 UTC 19 January 1989 with a central pressure of 1015 hPa, located approximately 250 km east of Nags Head, North Carolina. As with the IOP 4 cyclone, rapid intensification occurred immediately, with the central pressure falling to 982 hPa by 0000 UTC 20 January (1.87 Bergerons) with a location 400 km south-southeast of Sable Island, Nova Scotia. Subsequent intensification to 970 hPa occurred during the following 12 h.
e. Superstorm 1993
The well-developed surface low at 0600 UTC 12 March 1993 intensified rapidly during its passage though the Gulf of Mexico from 0600 UTC 12 to 0600 UTC 13 March. The manually analyzed (e.g., Kocin et al. 1995) central pressure fall of 25 hPa during this 24-h period is particularly impressive for the latitude, with a resultant latitude-adjusted 24-h intensification rate of 1.99 Bergerons. As noted by Dickinson et al. (1997), such an intensification was unprecedented at these latitudes.
f. Evaluation
The sea level pressure of the cyclones obtained from the manual tracking and from TRACK using each of the reanalyses are shown in Figs. 6–10. Additional information, including position discrepancies relative to the manual tracking are also shown for the reanalyses except for ERA5 – see below for information of this point (Tables 2 and 3). For the five cyclone histories, root-mean-square differences in position and intensity for the NOSMTH detections are 160–185 km and 5.7–8.5 hPa, respectively, which increase in SMTH to 196–208 km (12%–30%) and 14.9–15.7 hPa (76%–175%). We note that the highest resolution (CFSR) does not imply the greatest agreement in NOSMTH and interestingly, when smoothing is employed, it is most detrimental to the results obtained from the higher-resolution reanalyses. Crawford et al. (2021) find similarly for a different automated algorithm, stating “…using ERA5 sea level pressure fields at their finest-possible resolution does not necessarily lead to better cyclone detection and tracking” and “…cyclone frequency, life span, and average depth (are) insensitive to refining spatial resolution beyond 100 km.”
(top) Unsmoothed and (bottom) smoothed cyclone central pressure traces (hPa) for the Presidents’ Day Storm for each of the four reanalyses [(blue) CFSR, (purple) ERA5, (green) ERA-Interim, and (orange) NCEP]. Also shown is the pressure trace from the published analysis (black).
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 6, but for the Erica IOP 2 storm and showing the pressure trace from the Sanders analysis (black).
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 7, but for the Erica IOP 4 storm.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 7, but for the Erica IOP 5 storm.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 6, but for Superstorm ’93.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
Cyclone development for the five cyclones studied [Presidents’ Day storm (P-storm); ERICA IOP 2, IOP 4, and IOP 5; Superstorm 1993 (SS93)], as derived from manual tracking and TRACK. For TRACK, results for three analyses and for the unsmoothed (NOSMTH) and smoothed (SMTH) options are given. Max period denotes the beginning time of the 24-h period of maximum development, and B denotes the corresponding maximum development rate in Bergerons (see the text for details). The asterisk indicates that a 24-h development rate was not calculable for TRACK because of late detection, and in those instances B indicates the 12-h rate.
Root-mean-square difference of cyclone position (km) and central pressure (hPa) between manual tracking and TRACK with no smoothing (NOSMTH) and smoothing (SMTH) for the NCEP–DOE (NCEP), ERA-Interim (ERAI), and CFSR reanalyses for each of the five cyclones [Presidents’ Day storm (P-storm); ERICA IOP 2, IOP 4, and IOP 5; Superstorm 1993 (SS93)]. Total indicates the average of these measures over all five cases, but for each reanalysis separately.
Specifics regarding individual storms reveal similar patterns and are detailed below. In the case of the Presidents’ Day cyclone, the incipient cyclone’s closed circulation was well documented at 1800 UTC 18 February 1979 (Bosart 1981), but TRACK did not document its existence until considerably later (0600–1200 UTC 19 February), with the central pressure overestimated by 8 hPa in SMTH and 4 hPa in NOSMTH relative to the Bosart (1981) analysis. Similar differences in intensity continued through the remainder of the cyclone’s life cycle in NOSMTH, while the intensity differences grew to 18 hPa in SMTH (Fig. 6). Increased analysis resolution, even with smoothing turned off, did not resolve this problem (Fig. 6; Table 3).
To examine the possibility that none of the reanalyses considered here were sufficient for this case, we compared the ERA5 reanalysis with the manual analyses of Bosart (1981). The ERA5 analysis (not shown) does not represent the 1020-hPa mean sea level pressure contour at 1800 UTC 18 February 1979, instead representing this area with a broad-scale trough associated with the 1022-hPa isobar. By 0000 UTC 19 February, however, the closed contour is captured. Thus, in this case, using the ERA5 even without smoothing would not eliminate the late detection problem.
The algorithm fared better in the case of ERICA IOP 2, detecting this cyclone throughout its life cycle. NOSMTH closely approximated the storm’s intensity during its rapid deepening phase but underestimated the strength of the storm at its maximum intensity relative to the Sanders analysis (Fig. 7; Table 2). SMTH underestimated the storm’s intensity relative to the Sanders analysis throughout its life cycle. The maximum intensity differences in SMTH (∼20 hPa) were similar to those of the Presidents’ Day storm, but the maximum intensity differences in NOSMTH were substantially larger (10–15 hPa).
TRACK first detected the ERICA IOP 4 cyclone at 0600 UTC 4 January 1989, six hours later than the storm was detected manually. SMTH overestimated the initial central pressure relative to the Sanders analysis by ∼20 hPa, and NOSMTH overestimated the initial central pressure by ∼8 hPa. The NOSMTH algorithm’s detection of rapid intensification occurred only during the next 12 h, with intensification ending by 1800 UTC 4 January, despite the Sanders analysis showing intensification continuing until 0000 UTC 5 January (Fig. 8). In contrast, the SMTH algorithm indicated gradual intensification for a longer period than analyzed (until 1200 UTC 5 January). In all cases, the algorithm substantially underestimated the magnitude of the extraordinary intensification of this storm relative to the Sanders analysis (Table 2).
The first detection of the ERICA IOP 5 cyclone by the automated scheme occurred 12 h after the Sanders analysis, with rapid deepening documented only during the next 18-h period until 0600 UTC 20 January 1989 in NOSMTH (Fig. 9). The inconsistencies between the Sanders analysis and TRACK are similar to those seen for the ERICA IOP 4 cyclone, although the discrepancies in initial and maximum intensity are less than for the ERICA IOP 4 cyclone. NOSMTH provided better agreement with the maximum intensification derived from the Sanders analysis, but was still an underestimate relative to that manual analysis (Table 2).
For Superstorm 1993, the cyclone was detected by 0600 UTC 12 March 1993. In this event, TRACK overestimates the rate of intensification of the cyclone relative to the published analysis (Table 2), while the central pressure itself was also overestimated (by 5–10 hPa in NOSMTH and 10–15 hPa in SMTH throughout its life cycle), even during its passage into the more densely observed mid-Atlantic and northeastern United States (Fig. 10).
In summary, all cases except for the IOP 2 cyclone were detected at least 6 h after the first detection in the manual tracking. In all of these cases, the rapid intensification was proceeding either before, or at the time of the first detection by TRACK. We note that using TRACK with a 4-hPa anomaly minimum intensity threshold parameter might make it less sensitive to detecting early-stage cyclones. As a test, for the SMTH method, we reran the cyclone cases using a 2-hPa anomaly minimum intensity threshold parameter with the ERA-Interim reanalysis but found that the early-stage cyclones were still not detected by TRACK (not shown). This result might seem counterintuitive but results from the need to apply SMTH so as to avoid numerous spurious cyclone detections. The problem illustrated in Figs. 2–4 becomes more pronounced when using a 2-hPa threshold instead of a 4-hPa threshold.
The implication of this result is that analyses of crucial antecedent conditioning processes (e.g., Gyakum et al. 1992; Roebber 1993; Roebber et al. 1994) for these rapid cyclogeneses would be missed if a researcher were examining only the life cycles shown in TRACK. Furthermore, once detected, the cyclone’s intensity and intensification are consistently underestimated relative to existing manual tracking, even at later stages of the cyclone’s evolution, which are characterized by large-scale surface cyclonic circulations. It is also noteworthy that the cyclone position differences (Table 3) exclude those early-stage times at which TRACK does not yet recognize the existence of the storm. Thus, the average 160–210-km differences in cyclone position (i.e., ∼1.5°–2.0° latitude) represent the most favorable estimate of these discrepancies.
One should not conclude from these results that it is not possible to capture details of individual storms using TRACK. For example, using the ERA5 850-hPa vorticity as the input dataset, and retrieving the minimum sea level pressure from the unfiltered sea level pressure analysis, the maximum growth rates are comparable to those obtained from manual tracking, and TRACK detects the early stage of these cyclones at generally similar times to the manual tracking (not shown; anonymous reviewer personal communication). In this instance, however, it is not clear what advantage is obtained by using TRACK, because one could as readily manually track the cyclones directly from the reanalysis data, and such a procedure would not be feasible for a large sample such as those obtained from climate model simulations or to derive climatologies from a full reanalysis.
To explore this issue in greater detail, we conduct a direct comparison of all of the cyclone tracks obtained both manually and using TRACK during the time periods of the Presidents’ Day storm, ERICA IOP 2, ERICA IOP 4, and ERICA IOP 5. Here, we exclude the period for Superstorm 1993, because the manually analyzed maps are available only at 12-hourly intervals. For TRACK, we use the smoothed lower-tropospheric vorticity field obtained from the ERA5 reanalysis (see section 2 for further details on methods of vorticity tracking).
For the periods of the Presidents’ Day storm (Fig. 11), ERICA IOP 2 (Fig. 12), ERICA IOP 4 (Fig. 13) and ERICA IOP 5 (Fig. 14), the overabundance of cyclones detected by TRACK (bottom panels in each figure), relative to the manual analyses (top panels), is evident. The cyclone tracks (Figs. 11–14) for each of these four cases span the entire period from first to final detection in the manual analyses, in contrast to the shorter time spans of significant intensification discussed in the published analyses that are shown in Figs. 6–10. For ease of comparison, the top and bottom panels of the figures are longitudinally aligned. Considering the track segments in the top and bottom panels of each figure, it is evident that tracking using the relative vorticity field detects spurious cyclones that would not be detected by manual analysis. This is despite the fact that the vorticity field is smoothed prior to being input into TRACK and that ERA5 is the latest generation of reanalysis products. For the periods corresponding to ERICA IOP 2 and ERICA IOP 5, we see some remarkable TRACK cyclone lifespans that have no counterpart in the manual tracking data. Further, using TRACK with lower-tropospheric vorticity can produce displacements of the surface cyclone position from that obtained manually from the operational analysis (Figs. 11 and 15). In the case of the Presidents’ Day storm, TRACK appears to link the decaying cyclone with the north-south oriented surface trough and maintains continuity for another three days until 1800 UTC 27 February, presumably being associated with one or more additional distinct surface cyclones. The Presidents’ Day storm illustrates the challenges of producing automated detection and tracking at the earliest and latest stages of a cyclone’s life cycle. Considering the physics of surface cyclones and lower-tropospheric vorticity maxima, this lack of coupling may relate to variability in topographic features, hydrostatic stability, and friction, each of which can act to preclude a cyclone from persisting as long as lower-tropospheric vorticity maxima.
(top) The 6-hourly surface cyclone pressure center positions for the Presidents’ Day cyclone (Bosart analysis in cyan and operational manually analyzed positions, beginning at 1800 UTC 18 Feb 1979—hereinafter 1818) and all other surface cyclone positions detected from 1800 to 2512 within 20°–60°N and 40°–110°W. Red segments correspond to periods of cyclone intensification of at least one Bergeron. (bottom) The 6-hourly relative vorticity centers, from first to final detection, using TRACK with all detections occurring during the period of time common to that shown in the top panel. The cyan track is the Presidents’ Day cyclone. (top right) Sea level pressure field at 2418, with arrow showing the low, and its location on the manually analyzed track map.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 11, but for the Erica IOP 2 period (Sanders analysis) that includes the first detection period in the manual analysis from 0000 UTC 13 to 1200 UTC 17 Dec 1988.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 11, but for the Erica IOP 4 period (Sanders analysis) that includes the first detection period in the manual analysis from 0000 UTC 4 to 1200 UTC 9 Jan 1989.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
As in Fig. 11, but for the Erica IOP 5 period (Sanders analysis) that includes the first detection period in the manual analysis from 1200 UTC 18 to 1200 UTC 22 Jan 1989.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
ERA5 sea level pressure field (400-hPa interval), the position of the Presidents’ Day surface cyclone (L) as obtained from the operational manual analysis, and the position of the cyclone as obtained from TRACK (X) with the ERA5 lower-tropospheric vorticity at 1800 UTC 24 Feb 1979.
Citation: Monthly Weather Review 151, 11; 10.1175/MWR-D-22-0287.1
Next, in Table 4, we compare the times of initiation, maximum deepening, and dissipation and the magnitudes of maximum deepening for the four extreme storms obtained from manual tracking and TRACK, using the ERA5 smoothed lower-tropospheric vorticity and the corresponding sea level pressure values at the grid points identified by the vorticity maximum. TRACK first detected the cyclone-associated vorticity maximum with respect to the manually analyzed cyclone 6-h later (Presidents’ Day and IOP 4), at the same time (IOP 2), and 12-h later (IOP 5). TRACK last detected the cyclone-associated vorticity maximum with respect to the manually analyzed cyclone, 72-h later (Presidents’ Day), 36-h later (IOP 2), 6-h earlier (IOP 4), and 30-h later (IOP 5). In three of the four cases (IOP 2, IOP 4, IOP 5), the period of rapid deepening is delayed by 6-h relative to the manual tracking and the rapid development is weaker (Table 4), with RMSD of 0.80 Bergerons. Overall, even when accounting for the time offset, the RMSD of the position of the cyclone at the midpoint of maximum deepening is 126.9 km. Because the ERA5 has a grid spacing of 0.25°, these differences are substantial. For these four cyclones, the lifespans as obtained from TRACK are on average longer by 28.5 h, despite initiation being delayed by at least 6-h in three of the four events. Thus, we find substantial differences in core attributes of cyclone life histories for these important storms.
Cyclone histories for the Presidents’ Day storm (P-storm) and ERICA IOP 2, IOP 4, and IOP 5, as derived from manual tracking and TRACK. For TRACK, results are for the ERA5 reanalysis where the cyclone position is identified using the smoothed lower-tropospheric vorticity and the intensity is computed from the reanalysis sea level pressures at the location of the vorticity maximum. Here, S denotes the time of cyclone initiation, M denotes the beginning time of the 24-h period of maximum development, E denotes the end time of the cyclone identification, and B denotes the corresponding maximum development rate in Bergerons (see text for details).
4. Conclusions
Given the influence of midlatitude cyclones on weather and climate, it is of obvious interest to assure that methods developed to analyze their climatological characteristics in large datasets (such as now exist with reanalyses and climate model simulations) are adequate to the task. The size of a cyclone’s circulation, defined as the distance from its center to the nearest adjacent col, ridge, or anticyclone (e.g., Nielsen and Dole 1992), increases during the 24-h period of maximum intensification, and its often rather small size at the beginning time of maximum deepening might impede automated tracking.
As a first step in addressing this question, we examine the results obtained from a widely used cyclone tracking scheme (Hodges 1994) when applied to four reanalyses ranging from 0.5° to 2.5° horizontal grid spacing and varying levels of sophistication. We find substantial storm position and cyclone central pressure discrepancies from those obtained from detailed manual tracking of well-known cyclone histories available in the refereed literature. Importantly, these discrepancies are not eliminated by using higher-resolution reanalyses because spatial smoothing is needed to reduce spurious cyclone detections. This result aligns with other studies, including those using the improved reanalysis of ERA5 and a different tracking method (e.g., Crawford et al. 2021). Together, these results suggest that further improvements in reanalyses alone cannot solve these problems, most particularly for the generation of storm track climatologies in large datasets, where tracking settings cannot be customized for individual events.
Similar issues have been found in preliminary analyses of North Pacific cyclogenesis. For example, in the case of the rapidly deepening, eastern Pacific cyclone of 13–15 November 1981 (Reed and Albright 1986; Kuo and Reed 1988), the storm can be missed entirely by automated tracking. This can occur owing to the interaction between resolution and continuity requirements in a given scheme. At 0000 UTC 13 November 1981, the cyclone existed as a small but well-defined frontal wave, an existence that Reed and Albright (1986) document in their manual analyses using both ship observations and satellite imagery 12 h before. The smoothing necessitated by using higher-resolution reanalyses, or the intrinsic coarse-grained perspective of those at lower resolution, lead to a higher likelihood of an automated tracking scheme missing such a wave, as documented above in the North Atlantic for TRACK. Twelve hours later (1200 UTC 13 November 1981), when the wave becomes significantly more defined and is more visually prominent with a closed contour, there is a greater chance of detection by an automated tracking scheme. However, if the cyclone is initially missed, as we noted did occur in one such tracking scheme (not shown), detection continuity requirements can result in this storm never being recorded (e.g., if the portion of the cyclone track detected by the automated scheme did not extend for at least 1000 km or last for two days). In that instance, this well documented storm would be considered to never have existed.
Neu et al. (2013) have performed a comparison of 15 different automated tracking methods using a single analysis dataset. Although that work did not compare these automated climatologies with a manually derived “truth,” it is apparent from their results and those reported here that weaker and earlier stage cyclones pose a significant challenge to these efforts. Most of these situations involve smaller-scale secondary cyclogenesis, which when grown into larger scale circulations are then detected by the automated scheme. From a physical point of view, a complete understanding of cyclone development requires understanding of the early stages of cyclogenesis (e.g., Gyakum et al. 1992), something that is not likely to be interpreted correctly using current automated methods (see also Hewson and Titley 2010). These limitations also negatively impact the ability of these methods to properly interpret the full life cycle of the most extreme cyclones.
Because the large datasets available to modern researchers logistically limit the ability to revert to manual methods, the limitations of automated methods need to be understood in the context of the goals of a particular study. Here we have undertaken an analysis of only one automated scheme. We suggest that a coordinated effort, such as Neu et al. (2013) but applied to a set of well-studied cases as a baseline, is needed for a broader understanding of automated tracking limitations. For example, it would be useful to study the strengths and limitations of the University of Melbourne algorithm (Murray and Simmonds 1991; Simmonds and Keay 2000), which is intentionally designed to improve detection of wave cyclones and has been shown in some studies to improve upon the late detection bias (Mesquita et al. 2009; Neu et al. 2013).
Regardless, efforts to improve the ability of automated tracking methodologies to mimic the interpretation of trained analysts appear warranted. For example, it may be that insight into improved automated methods can be found in the extensive literature regarding displacement of particles in fluid flows [see Katz and Sheng (2010) for an overview of these methods].
We note that the extratropical cyclone tracking methods in use today are effectively based on synoptic heuristics and, as noted in the literature (e.g., Neu et al. 2013), provide variable results. Ullrich and Zarzycki (2017), recognizing this characteristic, have provided an open-source software framework that allows an ensemble approach to cyclone tracking. While this approach is justified given the inherent variability in current methods, Walker et al. (2020) argue that the diversity of methods “may lead to a lack of consensus on how ETC trends have been in the past, which makes it difficult to agree on how their numbers, intensities and impacts will change in the future.” We believe that the community can do better given the advent of highly capable pattern recognition algorithms such as convolutional neural networks (CNN). Because cyclone tracking is essentially a pattern recognition problem, we speculate that the application of CNNs, which have been highly successful across varying pattern recognition domains (e.g., speech translation: Abdel-Hamid et al. 2012; facial recognition: Singh et al. 2020), may provide superior automated schemes.
There are two ways to train machine learning algorithms: supervised and unsupervised learning. In the former case, one gives the algorithm a training dataset consisting of the inputs (e.g., the sea level pressure fields) and output data (e.g., the cyclones and tracks as obtained by an analyst) and seeks to develop the ability of the algorithm to match those outputs. In the unsupervised case, the algorithm is supplied only with the input data and the algorithm is tasked with finding patterns on its own. To our knowledge, supervised learning produces better results and does so more quickly. Thus, supervised machine learning requires well-curated datasets and while such cyclone tracking datasets derived from extensive manual tracking are available (e.g., Roebber 1984; Gyakum et al. 1989), each come with limitations. For example, in neither of the above datasets are global tracks available and while the Gyakum et al. (1989) spans a longer period, Roebber (1984) provides more spatial coverage, including continental cyclones. Accordingly, if such an effort is to be pursued, and one has less confidence in the output data from manual tracking or otherwise cannot justify the extensive effort required to assemble such data, a semisupervised approach might be employed. In semisupervised learning, a sample of labeled data is used to begin the training of the algorithm, and then the training is allowed to continue on the remaining (unlabeled) data. At this stage, because such efforts have yet to be undertaken, it is not clear if any of these approaches would be successful, but there appears to be sufficient reason to begin such efforts.
Acknowledgments.
This research was supported in part by the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant RGPIN-2016-04916. Yeechian Low provided some data and graphics support for this work. We thank the two anonymous reviewers and Sebastian Schemm for productive comments on an earlier version of this paper.
Data availability statement.
The reanalysis data used to perform this study are available online: NCEP Reanalysis 2 (http://doi.org/10.5065/KVQZ-YJ93), ERA-Interim (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim), CFSR (https://climatedataguide.ucar.edu/climate-data/climate-forecast-system-reanalysis-cfsr), CFSv2 (http://doi.org/10.5065/D69021ZF), and ERA5 (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5). The data generated using the tracking scheme can be provided on request, as can the Sanders manual analyses.
REFERENCES
Abdel-Hamid, O., A.-R. Mohamed, H. Jiang, and G. Penn, 2012: Applying convolutional neural networks concepts to hybrid NN-HMM model for speech recognition. 2012 IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Kyoto, Japan, Institute of Electrical and Electronics Engineers, 4277–4280, https://doi.org/10.1109/ICASSP.2012.6288864.
Alexander, G. D., J. A. Weinman, and J. L. Schols, 1998: The use of digital warping of microwave integrated water vapor imagery to improve forecasts of marine extratropical cyclones. Mon. Wea. Rev., 126, 1469–1496, https://doi.org/10.1175/1520-0493(1998)126<1469:TUODWO>2.0.CO;2.
Alexander, G. D., J. A. Weinman, V. M. Karyampudi, W. S. Olson, and A. C. L. Lee, 1999: The effect of assimilating rain rates derived from satellites and lightning on forecasts of the 1993 Superstorm. Mon. Wea. Rev., 127, 1433–1457, https://doi.org/10.1175/1520-0493(1999)127<1433:TEOARR>2.0.CO;2.
Blackmon, M. L., 1976: A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere. J. Atmos. Sci., 33, 1607–1623, https://doi.org/10.1175/1520-0469(1976)033<1607:ACSSOT>2.0.CO;2.
Blender, R., and M. Schubert, 2000: Cyclone tracking in different spatial and temporal resolutions. Mon. Wea. Rev., 128, 377–384, https://doi.org/10.1175/1520-0493(2000)128<0377:CTIDSA>2.0.CO;2.
Blender, R., K. Fraedrich, and F. Lunkeit, 1997: Identification of cyclone-track regimes in the North Atlantic. Quart. J. Roy. Meteor. Soc., 123, 727–741, https://doi.org/10.1002/qj.49712353910.
Blier, W., and R. M. Wakimoto, 1995: Observations of the early evolution of an explosive oceanic cyclone during ERICA IOP 5. Part I: Synoptic overview and mesoscale frontal structure. Mon. Wea. Rev., 123, 1288–1310, https://doi.org/10.1175/1520-0493(1995)123<1288:OOTEEO>2.0.CO;2.
Bosart, L. F., 1975: New England coastal frontogenesis. Quart. J. Roy. Meteor. Soc., 101, 957–978, https://doi.org/10.1002/qj.49710143016.
Bosart, L. F., 1981: The Presidents’ Day snowstorm of 18–19 February 1979: A subsynoptic-scale event. Mon. Wea. Rev., 109, 1542–1566, https://doi.org/10.1175/1520-0493(1981)109<1542:TPDSOF>2.0.CO;2.
Bosart, L. F., 1999: Observed cyclone life cycles. The Life Cycles of Extratropical Cyclones, M. A. Shapiro and S. Grønås, Eds., Amer. Meteor. Soc., 187–213.
Bosart, L. F., and S. C. Lin, 1984: A diagnostic analysis of the Presidents’ Day storm of February 1979. Mon. Wea. Rev., 112, 2148–2177, https://doi.org/10.1175/1520-0493(1984)112<2148:ADAOTP>2.0.CO;2.
Bosart, L. F., G. J. Hakim, K. R. Tyle, M. A. Bedrick, W. E. Bracken, M. J. Dickinson, and D. M. Schultz, 1996: Large-scale antecedent conditions associated with the 12–14 March 1993 cyclone (“Superstorm ’93”) over eastern North America. Mon. Wea. Rev., 124, 1865–1891, https://doi.org/10.1175/1520-0493(1996)124<1865:LSACAW>2.0.CO;2.
Bullock, T. A., and J. R. Gyakum, 1993: A diagnostic study of cyclogenesis in the western North Pacific Ocean. Mon. Wea. Rev., 121, 65–75, https://doi.org/10.1175/1520-0493(1993)121<0065:ADSOCI>2.0.CO;2.
Caplan, P. M., 1995: The 12–14 March 1993 Superstorm: Performance of the NMC global medium-range model. Bull. Amer. Meteor. Soc., 76, 201–212, https://doi.org/10.1175/1520-0477(1995)076<0201:TMSPOT>2.0.CO;2.
Chang, E. K. M., S. Lee, and K. L. Swanson, 2002: Storm track dynamics. J. Climate, 15, 2163–2183, https://doi.org/10.1175/1520-0442(2002)015<02163:STD>2.0.CO;2.
Chang, E. K. M., Y. Guo, and X. Xia, 2012: CMIP5 multimodel ensemble projection of storm track change under global warming. J. Geophys. Res., 117, D23118, https://doi.org/10.1029/2012JD018578.
Chang, S. W., and T. R. Holt, 1994: Impact of assimilating SSM/I rainfall rates on numerical prediction of winter cyclones. Mon. Wea. Rev., 122, 151–164, https://doi.org/10.1175/1520-0493(1994)122<0151:IOASRR>2.0.CO;2.
Chang, S. W., R. J. Alliss, S. Raman, and J.-J. Shi, 1993: SSM/I observations of ERICA IOP 4 marine cyclone: A comparison with in situ observations and model simulation. Mon. Wea. Rev., 121, 2452–2464, https://doi.org/10.1175/1520-0493(1993)121<2452:SOOEIM>2.0.CO;2.
Chang, S. W., T. R. Holt, and K. D. Sashegyi, 1996: A numerical study of the ERICA IOP 4 marine cyclone. Mon. Wea. Rev., 124, 27–46, https://doi.org/10.1175/1520-0493(1996)124<0027:ANSOTE>2.0.CO;2.
Cohen, R. A., and C. W. Kreitzberg, 1997: Airstream boundaries in numerical weather simulations. Mon. Wea. Rev., 125, 168–183, https://doi.org/10.1175/1520-0493(1997)125<0168:ABINWS>2.0.CO;2.
Colucci, S. J., 1976: Winter cyclone frequencies over the eastern United States and adjacent western Atlantic, 1964–1973. Bull. Amer. Meteor. Soc., 57, 548–553, https://doi.org/10.1175/1520-0477(1976)057<0548:WCFOTE>2.0.CO;2.
Crawford, A. D., E. A. P. Schreiber, N. Sommer, M. C. Serreze, J. C. Stroeve, and D. G. Barber, 2021: Sensitivity of Northern Hemisphere cyclone detection and tracking results to fine spatial and temporal resolution using ERA5. Mon. Wea. Rev., 149, 2581–2598, https://doi.org/10.1175/MWR-D-20-0417.1.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Desjardins, S., R. Benoit, and V. Swail, 1998: The influence of mesoscale features of the sea surface temperature distribution on marine boundary layer winds off the Scotian shelf during the Superstorm of March 1993. Mon. Wea. Rev., 126, 2793–2808, https://doi.org/10.1175/1520-0493(1998)126<2793:TIOMFO>2.0.CO;2.
Dickinson, M. J., L. F. Bosart, W. E. Bracken, G. J. Hakim, D. M. Schultz, M. A. Bedrick, and K. R. Tyle, 1997: The March 1993 Superstorm cyclogenesis: Incipient phase synoptic- and convective-scale flow interaction and model performance. Mon. Wea. Rev., 125, 3041–3072, https://doi.org/10.1175/1520-0493(1997)125<3041:TMSCIP>2.0.CO;2.
Doblas-Reyes, F. J., and M. Déqué, 1998: A flexible bandpass filter design procedure applied to midlatitude intraseasonal variability. Mon. Wea. Rev., 126, 3326–3335, https://doi.org/10.1175/1520-0493(1998)126<3326:AFBFDP>2.0.CO;2.
Gilhousen, D. B., 1994: The value of NDBC observations during March 1993’s “storm of the century.” Wea. Forecasting, 9, 255–264, https://doi.org/10.1175/1520-0434(1994)009<0255:TVONOD>2.0.CO;2.
Grise, K. M., S.-W. Son, and J. R. Gyakum, 2013: Intraseasonal and interannual variability in North American storm tracks and its relationship to equatorial Pacific variability. Mon. Wea. Rev., 141, 3610–3625, https://doi.org/10.1175/MWR-D-12-00322.1.
Grotjahn, R., 1993: Global Atmospheric Circulations: Observations and Theories. Oxford University Press, 430 pp.
Gulev, S. K., O. Zolina, and S. Grigoriev, 2001: Extratropical cyclone variability in the Northern Hemisphere winter from the NCEP/NCAR reanalysis data. Climate Dyn., 17, 795–809, https://doi.org/10.1007/s003820000145.
Gyakum, J. R., 1991: Meteorological precursors to the explosive intensification of the QE II storm. Mon. Wea. Rev., 119, 1105–1131, https://doi.org/10.1175/1520-0493(1991)119<1105:MPTTEI>2.0.CO;2.
Gyakum, J. R., and R. E. Danielson, 2000: Analysis of meteorological precursors to ordinary and explosive cyclogenesis in the western North Pacific. Mon. Wea. Rev., 128, 851–863, https://doi.org/10.1175/1520-0493(2000)128<0851:AOMPTO>2.0.CO;2.
Gyakum, J. R., J. R. Anderson, R. H. Grumm, and E. L. Gruner, 1989: North Pacific cold-season surface cyclone activity: 1975–1983. Mon. Wea. Rev., 117, 1141–1155, https://doi.org/10.1175/1520-0493(1989)117<1141:NPCSSC>2.0.CO;2.
Gyakum, J. R., P. J. Roebber, and T. Bullock, 1992: The role of antecedent surface vorticity development as a conditioning process in explosive cyclone intensification. Mon. Wea. Rev., 120, 1465–1489, https://doi.org/10.1175/1520-0493(1992)120<1465:TROASV>2.0.CO;2.
Hadlock, R., and C. W. Kreitzberg, 1988: The Experiment on Rapidly Intensifying Cyclones over the Atlantic (ERICA) field study: Objectives and plans. Bull. Amer. Meteor. Soc., 69, 1309–1320, https://doi.org/10.1175/1520-0477(1988)069<1309:TEORIC>2.0.CO;2.
Hersbach, H., and Coauthors, 2020: The ERA 5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hewson, T. D., and H. A. Titley, 2010: Objective identification, typing and tracking of the complete life-cycles of cyclonic features at high spatial resolution. Meteor. Appl., 17, 355–381, https://doi.org/10.1002/met.204.
Hodges, K. I., 1994: A general method for tracking analysis and its application to meteorological data. Mon. Wea. Rev., 122, 2573–2586, https://doi.org/10.1175/1520-0493(1994)122<2573:AGMFTA>2.0.CO;2.
Hodges, K. I., 1995: Feature tracking on the unit sphere. Mon. Wea. Rev., 123, 3458–3465, https://doi.org/10.1175/1520-0493(1995)123<3458:FTOTUS>2.0.CO;2.
Hodges, K. I., 1999: Adaptive constraints for feature tracking. Mon. Wea. Rev., 127, 1362–1373, https://doi.org/10.1175/1520-0493(1999)127<1362:ACFFT>2.0.CO;2.
Hodges, K. I., R. W. Lee, and L. Bengtsson, 2011: A comparison of extratropical cyclones in recent reanalyses ERA-Interim, NASA MERRA, NCEP CFSR, and JRA-25. J. Climate, 24, 4888–4906, https://doi.org/10.1175/2011JCLI4097.1.
Hoskins, B. J., and K. I. Hodges, 2002: New perspectives on the Northern Hemisphere winter storm tracks. J. Atmos. Sci., 59, 1041–1061, https://doi.org/10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2.
Huo, Z., D.-L. Zhang, J. R. Gyakum, and A. Staniforth, 1995: A diagnostic analysis of the Superstorm of March 1993. Mon. Wea. Rev., 123, 1740–1761, https://doi.org/10.1175/1520-0493(1995)123<1740:ADAOTS>2.0.CO;2.
Huo, Z., D.-L. Zhang, and J. R. Gyakum, 1998: An application of potential vorticity inversion to improving the numerical prediction of the March 1993 Superstorm. Mon. Wea. Rev., 126, 424–436, https://doi.org/10.1175/1520-0493(1998)126<0424:AAOPVI>2.0.CO;2.
Jung, T., and Coauthors, 2012: High-resolution global climate simulations with the ECMWF model in Project Athena: Experimental design, model climate, and seasonal forecast skill. J. Climate, 25, 3155–3172, https://doi.org/10.1175/JCLI-D-11-00265.1.
Kanamitsu, M., W. Ebisuzaki, J. Woolen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1644, https://doi.org/10.1175/BAMS-83-11-1631.
Katz, J., and J. Sheng, 2010: Applications of holography in fluid mechanics and particle dynamics. Annu. Rev. Fluid Mech., 42, 531–555, https://doi.org/10.1146/annurev-fluid-121108-145508.
Klein, W. H., 1957: Principal tracks and mean frequencies of cyclones and anticyclones in the Northern Hemisphere. U.S. Weather Bureau Research Paper 40, 60 pp.
Kocin, P. J., P. N. Schumacher, R. F. Morales Jr., and L. W. Uccellini, 1995: Overview of the 12–14 March 1993 Superstorm. Bull. Amer. Meteor. Soc., 76, 165–182, https://doi.org/10.1175/1520-0477(1995)076<0165:OOTMS>2.0.CO;2.
König, W., R. Sausen, and F. Sielmann, 1993: Objective identification of cyclones in GCM simulations. J. Climate, 6, 2217–2231, https://doi.org/10.1175/1520-0442(1993)006<2217:OIOCIG>2.0.CO;2.
Kuo, Y.-H., and R. J. Reed, 1988: Numerical simulation of an explosively deepening cyclone in the eastern Pacific. Mon. Wea. Rev., 116, 2081–2105, https://doi.org/10.1175/1520-0493(1988)116<2081:NSOAED>2.0.CO;2.
Kuo, Y.-H., R. J. Reed, and Y. Liu, 1996: The ERICA IOP 5 storm. Part III: Mesoscale cyclogenesis and precipitation parameterization. Mon. Wea. Rev., 124, 1409–1434, https://doi.org/10.1175/1520-0493(1996)124<1409:TEISPI>2.0.CO;2.
Lackmann, G. M., D. Keyser, and L. F. Bosart, 1997: A characteristic life cycle of upper-tropospheric cyclogenetic precursors during the Experiment on Rapidly Intensifying Cyclones over the Atlantic (ERICA). Mon. Wea. Rev., 125, 2729–2758, https://doi.org/10.1175/1520-0493(1997)125<2729:ACLCOU>2.0.CO;2.
Lackmann, G. M., D. Keyser, and L. F. Bosart, 1999: Energetics of an intensifying jet streak during the Experiment on Rapidly Intensifying Cyclones over the Atlantic (ERICA). Mon. Wea. Rev., 127, 2777–2795, https://doi.org/10.1175/1520-0493(1999)127<2777:EOAIJS>2.0.CO;2.
Liu, C.-H., R. M. Wakimoto, and F. Roux, 1997: Observations of mesoscale circulations within extratropical cyclones over the North Atlantic Ocean during ERICA. Mon. Wea. Rev., 125, 341–364, https://doi.org/10.1175/1520-0493(1997)125<0341:OOMCWE>2.0.CO;2.
Manobianco, J., S. Koch, V. M. Karyampudi, and A. J. Negri, 1994: The impact of assimilating satellite-derived precipitation rates on numerical simulations of the ERICA IOP 4 cyclone. Mon. Wea. Rev., 122, 341–365, https://doi.org/10.1175/1520-0493(1994)122<0341:TIOASD>2.0.CO;2.
Mesquita, M. D. S., D. E. Atkinson, I. Simmonds, K. Keay, and J. Gottschalck, 2009: New perspectives on the synoptic development of the severe October 1992 Nome storm. Geophys. Res. Lett., 36, L13808, https://doi.org/10.1029/2009GL038824.
Murray, R. J., and I. Simmonds, 1991: A numerical scheme for tracking cyclone centres from digital data, Part I: Development and operation of the scheme. Aust. Meteor. Mag., 39, 155–166.
Narita, D., R. S. J. Tol, and D. Anthoff, 2010: Economic costs of extratropical storms under climate change: An application of FUND. J. Environ. Plann. Manage., 53, 371–384, https://doi.org/10.1080/09640561003613138.
Neu, U., and Coauthors, 2013: IMILAST: A community effort to intercompare extratropical cyclone detection and tracking algorithms. Bull. Amer. Meteor. Soc., 94, 529–547, https://doi.org/10.1175/BAMS-D-11-00154.1.
Nielsen, J. W., and R. M. Dole, 1992: A survey of extratropical cyclone characteristics during GALE. Mon. Wea. Rev., 120, 1156–1168, https://doi.org/10.1175/1520-0493(1992)120<1156:ASOECC>2.0.CO;2.
Orlanski, I., and J. Sheldon, 1993: A case of downstream baroclinic development over western North America. Mon. Wea. Rev., 121, 2929–2950, https://doi.org/10.1175/1520-0493(1993)121<2929:ACODBD>2.0.CO;2.
Parsons, K. E., and P. J. Smith, 2004: An investigation of extratropical cyclone development using a scale-separation technique. Mon. Wea. Rev., 132, 956–974, https://doi.org/10.1175/1520-0493(2004)132<0956:AIOECD>2.0.CO;2.
Pauley, P. M., and B. J. Bramer, 1992: The effect of resolution on the depiction of central pressure for an intense oceanic extratropical cyclone. Mon. Wea. Rev., 120, 757–769, https://doi.org/10.1175/1520-0493(1992)120<0757:TEOROT>2.0.CO;2.
Penny, S., G. H. Roe, and D. S. Battisti, 2010: The source of the midwinter suppression in storminess over the North Pacific. J. Climate, 23, 634–648, https://doi.org/10.1175/2009JCLI2904.1.
Petterssen, S., 1956: Motion and Motion Systems. Vol. I, Weather Analysis and Forecasting, McGraw-Hill, 428 pp.
Priestley, M. D. K., D. Ackerley, J. L. Catto, K. I. Hodges, R. E. McDonald, and R. W. Lee, 2020: An overview of the extratropical storm tracks in CMIP6 historical simulations. J. Climate, 33, 6315–6343, https://doi.org/10.1175/JCLI-D-19-0928.1.
Raible, C. C., P. M. Della-Marta, C. Schwierz, H. Wernli, and R. Blender, 2008: Northern Hemisphere extratropical cyclones: A comparison of detection and tracking methods and different reanalyses. Mon. Wea. Rev., 136, 880–897, https://doi.org/10.1175/2007MWR2143.1.
Rausch, R. L. M., and P. J. Smith, 1996: A diagnosis of a model-simulated explosively developing extratropical cyclone. Mon. Wea. Rev., 124, 875–904, https://doi.org/10.1175/1520-0493(1996)124<0875:ADOAMS>2.0.CO;2.
Reed, R. J., and M. D. Albright, 1986: A case study of explosive cyclogenesis in the eastern Pacific. Mon. Wea. Rev., 114, 2297–2319, https://doi.org/10.1175/1520-0493(1986)114<2297:ACSOEC>2.0.CO;2.
Reed, R. J., G. A. Grell, and Y.-H. Kuo, 1993a: The ERICA IOP 5 storm. Part I: Analysis and simulation. Mon. Wea. Rev., 121, 1577–1594, https://doi.org/10.1175/1520-0493(1993)121<1577:TEISPI>2.0.CO;2.
Reed, R. J., G. A. Grell, and Y.-H. Kuo, 1993b: The ERICA IOP 5 storm. Part II: Sensitivity tests and further diagnosis based on model output. Mon. Wea. Rev., 121, 1595–1612, https://doi.org/10.1175/1520-0493(1993)121<1595:TEISPI>2.0.CO;2.
Reed, R. J., Y.-H. Kuo, and S. Low-Nam, 1994: An adiabatic simulation of the ERICA IOP 4 storm: An example of quasi-ideal frontal cyclone development. Mon. Wea. Rev., 122, 2688–2708, https://doi.org/10.1175/1520-0493(1994)122<2688:AASOTE>2.0.CO;2.
Reitan, C. H., 1974: Frequencies of cyclones and cyclogenesis for North America, 1951–1970. Mon. Wea. Rev., 102, 861–868, https://doi.org/10.1175/1520-0493(1974)102<0861:FOCACF>2.0.CO;2.
Roebber, P. J., 1984: Statistical analysis and updated climatology of explosive cyclones. Mon. Wea. Rev., 112, 1577–1589, https://doi.org/10.1175/1520-0493(1984)112<1577:SAAUCO>2.0.CO;2.
Roebber, P. J., 1989: On the statistical analysis of cyclone deepening rates. Mon. Wea. Rev., 117, 2293–2298, https://doi.org/10.1175/1520-0493(1989)117<2293:OTSAOC>2.0.CO;2.
Roebber, P. J., 1993: A diagnostic case study of self-development as an antecedent conditioning process in explosive cyclogenesis. Mon. Wea. Rev., 121, 976–1006, https://doi.org/10.1175/1520-0493(1993)121<0976:ADCSOS>2.0.CO;2.
Roebber, P. J., and M. R. Schumann, 2011: Physical processes governing the rapid deepening tail of maritime cyclogenesis. Mon. Wea. Rev., 139, 2776–2789, https://doi.org/10.1175/MWR-D-10-05002.1.
Roebber, P. J., J. R. Gyakum, and D. N. Trat, 1994: Coastal frontogenesis and precipitation during ERICA IOP 2. Wea. Forecasting, 9, 21–44, https://doi.org/10.1175/1520-0434(1994)009<0021:CFAPDE>2.0.CO;2.
Rohrer, M., O. Martius, C. C. Raible, and S. Brönnimann, 2020: Sensitivity of blocks and cyclones in ERA5 to spatial resolution and definition. Geophys. Res. Lett., 47, e2019GL085582, https://doi.org/10.1029/2019GL085582.
Rudeva, I., S. K. Gulev, I. Simmonds, and N. Tilinina, 2014: The sensitivity of characteristics of cyclone activity to identification procedures in tracking algorithms. Tellus, 66A, 24961, https://doi.org/10.3402/tellusa.v66.24961.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057, https://doi.org/10.1175/2010BAMS3001.1.
Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 2185–2208, https://doi.org/10.1175/JCLI-D-12-00823.1.
Sanders, F., 1990: Surface analysis over the oceans—Searching for sea truth. Wea. Forecasting, 5, 596–612, https://doi.org/10.1175/1520-0434(1990)005<0596:SAOTOF>2.0.CO;2.
Sanders, F., and J. R. Gyakum, 1980: Synoptic-dynamic climatology of the “bomb.” Mon. Wea. Rev., 108, 1589–1606, https://doi.org/10.1175/1520-0493(1980)108<1589:SDCOT>2.0.CO;2.
Sanders, F., and C. A. Doswell III, 1995: A case for detailed surface analysis. Bull. Amer. Meteor. Soc., 76, 505–521, https://doi.org/10.1175/1520-0477(1995)076<0505:ACFDSA>2.0.CO;2.
Schemm, S., M. Sprenger, and H. Wernli, 2018: When during their life cycle are extratropical cyclones attended by fronts? Bull. Amer. Meteor. Soc., 99, 149–165, https://doi.org/10.1175/BAMS-D-16-0261.1.
Schultz, D. M., and W. J. Steenburgh, 1999: The formation of a forward-tilting cold front with multiple cloud bands during Superstorm 1993. Mon. Wea. Rev., 127, 1108–1124, https://doi.org/10.1175/1520-0493(1999)127<1108:TFOAFT>2.0.CO;2.
Schultz, D. M., W. E. Bracken, L. F. Bosart, G. J. Hakim, M. A. Bedrick, M. J. Dickinson, and K. R. Tyle, 1997: The 1993 Superstorm cold surge: Frontal structure, gap flow, and tropical impact. Mon. Wea. Rev., 125, 5–39, https://doi.org/10.1175/1520-0493(1997)125<0005:TSCSFS>2.0.CO;2.
Schultz, D. M., and Coauthors, 2019: Extratropical cyclones: A century of research on meteorology’s centerpiece. A Century of Progress in Atmospheric and Related Sciences: Celebrating the American Meteorological Society Centennial, Meteor. Monogr., No. 59, Amer. Meteor. Soc., https://doi.org/10.1175/AMSMONOGRAPHS-D-18-0015.1.
Shaw, T. A., and Coauthors, 2016: Storm track processes and the opposing influences of climate change. Nat. Geosci., 9, 656–664, https://doi.org/10.1038/ngeo2783.
Simmonds, I., and K. Keay, 2000: Mean Southern Hemisphere extratropical cyclone behavior in the 40-Year NCEP–NCAR reanalysis. J. Climate, 13, 873–885, https://doi.org/10.1175/1520-0442(2000)013<0873:MSHECB>2.0.CO;2.
Simmonds, I., and I. Rudeva, 2014: A comparison of tracking methods for extreme cyclones in the Arctic basin. Tellus, 66A, 25252, https://doi.org/10.3402/tellusa.v66.25252.
Simmonds, I., K. Keay, and J. A. T. Bye, 2012: Identification and climatology of Southern Hemisphere mobile fronts in a modern reanalysis. J. Climate, 25, 1945–1962, https://doi.org/10.1175/JCLI-D-11-00100.1.
Simmons, A. J., and B. J. Hoskins, 1979: The downstream and upstream development of unstable baroclinic waves. J. Atmos. Sci., 36, 1239–1254, https://doi.org/10.1175/1520-0469(1979)036<1239:TDAUDO>2.0.CO;2.
Sinclair, M. R., 1994: An objective cyclone climatology for the Southern Hemisphere. Mon. Wea. Rev., 122, 2239–2256, https://doi.org/10.1175/1520-0493(1994)122<2239:AOCCFT>2.0.CO;2.
Sinclair, M. R., 1997: Objective identification of cyclones and their circulation intensity, and climatology. Wea. Forecasting, 12, 595–612, https://doi.org/10.1175/1520-0434(1997)012<0595:OIOCAT>2.0.CO;2.
Singh, N. S., S. Hariharan, and M. Gupta, 2020: Facial recognition using deep learning. Advances in Data Sciences, Security and Applications, V. Jain et al., Eds., Lecture Notes in Electrical Engineering, Vol. 612, Springer, 375–382, https://doi.org/10.1007/978-981-15-0372-6_30.
Swanson, K. L., 2007: Storm track dynamics. The Global Circulation of the Atmosphere, T. Schneider, Ed., Princeton University Press, 78–108.
Swanson, K. L., and P. J. Roebber, 2008: The impact of analysis error on medium-range weather forecasts. Mon. Wea. Rev., 136, 3425–3431, https://doi.org/10.1175/2008MWR2475.1.
Trigo, I. F., 2006: Climatology and interannual variability of storm-tracks in the Euro-Atlantic sector: A comparison between ERA-40 and NCEP/NCAR reanalyses. Climate Dyn., 26, 127–143, https://doi.org/10.1007/s00382-005-0065-9.
Uccellini, L. W., 1990: Processes contributing to the rapid development of extratropical cyclones. Extratropical Cyclones: The Erik Palmén Memorial Volume, C.W. Newton and E. Holopainen, Eds., Amer. Meteor. Soc., 81–105.
Uccellini, L. W., P. J. Kocin, R. A. Petersen, C. H. Wash, and K. F. Brill, 1984: The Presidents’ Day cyclone of 18–19 February 1979: Synoptic overview and analysis of the subtropical jet streak influencing the pre-cyclogenetic period. Mon. Wea. Rev., 112, 31–55, https://doi.org/10.1175/1520-0493(1984)112<0031:TPDCOF>2.0.CO;2.
Uccellini, L. W., D. Keyser, K. F. Brill, and C. H. Wash, 1985: The Presidents’ Day cyclone of 18–19 February 1979: Influence of upstream trough amplification and associated tropopause folding on rapid cyclogenesis. Mon. Wea. Rev., 113, 962–988, https://doi.org/10.1175/1520-0493(1985)113<0962:TPDCOF>2.0.CO;2.
Uccellini, L. W., R. A. Petersen, P. J. Kocin, K. F. Brill, and J. J. Tuccillo, 1987: Synergistic interactions between an upper-level jet streak and diabatic processes that influence the development of a low-level jet and a secondary coastal cyclone. Mon. Wea. Rev., 115, 2227–2261, https://doi.org/10.1175/1520-0493(1987)115<2227:SIBAUL>2.0.CO;2.
Uccellini, L. W., S. F. Corfidi, N. W. Junker, P. J. Kocin, and D. A. Olson, 1992: Report on the surface analysis workshop at the National Meteorological Center 25–28 March 1991. Bull. Amer. Meteor. Soc., 73, 459–476, https://doi.org/10.1175/1520-0477-73.4.459.
Uccellini, L. W., P. J. Kocin, R. S. Schneider, P. M. Stokols, and R. A. Dorr, 1995: Forecasting the 12–14 March 1993 Superstorm. Bull. Amer. Meteor. Soc., 76, 183–200, https://doi.org/10.1175/1520-0477(1995)076<0183:FTMS>2.0.CO;2.
Ulbrich, U., G. C. Leckebusch, and J. G. Pinto, 2009: Extra-tropical cyclones in the present and future climate: A review. Theor. Appl. Climatol., 96, 117–131, https://doi.org/10.1007/s00704-008-0083-8.
Ullrich, P. A., and C. M. Zarzycki, 2017: TempestExtremes: A framework for scale-insensitive pointwise feature tracking on unstructured grids. Geosci. Model Dev., 10, 1069–1090, https://doi.org/10.5194/gmd-10-1069-2017.
Vessey, A. F., K. I. Hodges, L. C. Shaffrey, and J. J. Day, 2020: An inter-comparison of Arctic synoptic scale storms between four global reanalysis datasets. Climate Dyn., 54, 2777–2795, https://doi.org/10.1007/s00382-020-05142-4.
Wakimoto, R. M., W. Blier, and C. Liu, 1992: The frontal structure of an explosive oceanic cyclone: Airborne radar observations of ERICA IOP 4. Mon. Wea. Rev., 120, 1135–1155, https://doi.org/10.1175/1520-0493(1992)120<1135:TFSOAE>2.0.CO;2.
Wakimoto, R. M., N. T. Atkins, and C. Liu, 1995: Observations of the early evolution of an explosive oceanic cyclone during ERICA IOP 5. Part II: Airborne Doppler analysis of the mesoscale circulation and frontal structure. Mon. Wea. Rev., 123, 1311–1327, https://doi.org/10.1175/1520-0493(1995)123<1311:OOTEEO>2.0.CO;2.
Walker, E., D. Mitchell, and W. Seviour, 2020: The numerous approaches to tracking extratropical cyclones and the challenges they present. Weather, 75, 336–341, https://doi.org/10.1002/wea.3861.
Wallace, J. M., G.-H. Lim, and M. L. Blackmon, 1988: Relationship between cyclone tracks, anticyclone tracks, and baroclinic waveguides. J. Atmos. Sci., 45, 439–462, https://doi.org/10.1175/1520-0469(1988)045<0439:RBCTAT>2.0.CO;2.
Wernli, H., and C. Schwierz, 2006: Surface cyclones in the ERA-40 dataset (1958–2001). Part I: Novel identification method and global climatology. J. Atmos. Sci., 63, 2486–2507, https://doi.org/10.1175/JAS3766.1.
Whitaker, J. S., L. W. Uccellini, and K. F. Brill, 1988: A model-based diagnostic study of the rapid development phase of the Presidents’ Day cyclone. Mon. Wea. Rev., 116, 2337–2365, https://doi.org/10.1175/1520-0493(1988)116<2337:AMBDSO>2.0.CO;2.
Whittaker, L. M., and L. H. Horn, 1981: Geographical and seasonal distribution of North American cyclogenesis, 1958–1977. Mon. Wea. Rev., 109, 2312–2322, https://doi.org/10.1175/1520-0493(1981)109<2312:GASDON>2.0.CO;2.
Xiao, Q., X. Zou, and Y.-H. Kuo, 2000: Incorporating the SSM/I-derived precipitable water and rainfall rate into a numerical model: A case study for the ERICA IOP-4 cyclone. Mon. Wea. Rev., 128, 87–108, https://doi.org/10.1175/1520-0493(2000)128<0087:ITSIDP>2.0.CO;2.
Xiao, Q., X. Zou, M. Pondeca, M. A. Shapiro, and C. Velden, 2002: Impact of GMS-5 and GOES-9 satellite-derived winds on the prediction of a NORPEX extratropical cyclone. Mon. Wea. Rev., 130, 507–528, https://doi.org/10.1175/1520-0493(2002)130<0507:IOGAGS>2.0.CO;2.
Zhu, H., and A. Thorpe, 2006: Predictability of extratropical cyclones: The influence of initial condition and model uncertainties. J. Atmos. Sci., 63, 1483–1497, https://doi.org/10.1175/JAS3688.1.
Zishka, K. M., and P. J. Smith, 1980: The climatology of cyclones and anticyclones over North America and surrounding ocean environs for January and July, 1950–77. Mon. Wea. Rev., 108, 387–401, https://doi.org/10.1175/1520-0493(1980)108<0387:TCOCAA>2.0.CO;2.
Zou, X., Y.-H. Kuo, and S. Low-Nam, 1998: Medium-range prediction of an extratropical oceanic cyclone: Impact of initial state. Mon. Wea. Rev., 126, 2737–2763, https://doi.org/10.1175/1520-0493(1998)126<2737:MRPOAE>2.0.CO;2.
