Bermuda Subtropical Storms

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/225858605

Article in Meteorology and Atmospheric Physics · August 2007 DOI: 10.1007/s00703-006-0255-y

CITATIONS READS 13 33

5 authors, including:

Mark Guishard Elizabeth Harris Bermuda Institute of Ocean Sciences Ariel Re

18 PUBLICATIONS 174 CITATIONS 2 PUBLICATIONS 24 CITATIONS

SEE PROFILE SEE PROFILE

Jenni L Evans Pennsylvania State University

75 PUBLICATIONS 3,607 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Interaction of MCSs coupled to AEWs over North Africa View project

Risk Prediction Initiative Report View project

All content following this page was uploaded by Jenni L Evans on 10 July 2018.

The user has requested enhancement of the downloaded file. Meteorol Atmos Phys (2007) DOI 10.1007/s00703-006-0255-y Printed in The Netherlands

1 Bermuda Weather Service, BAS-Serco Ltd., St. Georges, Bermuda 2 Department of Meteorology, The Pennsylvania State University, PA, USA 3 Department of Meteorology, Florida State University, FL, USA

Bermuda subtropical storms

M. P. Guishard1, E. A. Nelson1, J. L. Evans2, R. E. Hart3, and D. G. O’Connell1

With 14 Figures

Received August 4, 2006; accepted September 19, 2006 Published online: March 14, 2007 # Springer-Verlag 2007

Summary customers (signiﬁcant for an island with approx- 1 This investigation focuses on North Atlantic subtropical cy- imately 62,000 inhabitants ). This storm was clones which tracked within 100 nautical miles (185 km) of determined to have been a ‘‘subtropical storm’’ Bermuda from 1957 to 2005, identiﬁed through subtropi- by the US National Hurricane Center (NHC), cal structural characteristics distinguished using Cyclone the regional tropical cyclone authority under the Phase Space, from the European Centre for Medium-Range World Meteorological Organisation (WMO). The Weather Forecasts 45-year reanalyses. The study assesses the characteristics of these hybrid storms that affect the storm continued to evolve more tropical charac- Island, in order to aid the local forecaster. Reanalysis charts, teristics and, on October 12th, the storm was surface analyses, local observations, HURDAT tracks, and named Karen. Karen strengthened to a Category- satellite pictures, where available, were examined. This 2 hurricane, and went on to strike Nova Scotia data shows that subtropical cyclones affecting Bermuda as a tropical storm. During the approach and usually form in close proximity, to the south-southwest, rapid evolution of Karen over Bermuda, a Gale over water of an average of 26 C, under moderate vertical À1 wind shear, with an upper trough lying to the west-north- Warning (for winds of at least 17 m s affecting west. They then move in a north-northeastward direction, the local and marine area) was issued and later intensifying quickly, but not often reaching a peak intensity upgraded to a Storm Warning (25 m sÀ1)by of more than 26 m sÀ1. They generally have their begin- the Bermuda Weather Service (BWS). The gale- nings along old baroclinic zones. September is the peak to hurricane-force winds, warm lower core and month of occurrence. A direct hit by a severe subtropical cyclone, producing locally observed winds of over 26 m sÀ1, widespread (albeit asymmetric) convection asso- appears to be a rare event. However, these storms are cer- ciated with the system (Fig. 1) led forecasters to tainly a threat to the Island, particularly due to their lack of conclude that this system had at least partial trop- predictability, and conditions conducive to an incipient sub- ical characteristics (Williams, 2002). However, tropical cyclone with potential to affect the Island should the absence of advisories from the US National always be closely monitored. Hurricane Center (NHC) during the onset of gale force winds restricted local forecasters to the 1. Introduction recorded issuance of local marine wind warnings. The ensuing media coverage over the next two In October 2001, a rapidly developing marine storm produced hurricane-force winds over 1 From Government of Bermuda Department of Statistics Census Bermuda, resulting in the loss of power to 23,000 2000. M. P. Guishard et al

other than those already set forth in Guishard (2006); and (c) determine the conditions under which Bermuda STs occur, thus providing a basis for the successful forecasting of such systems.

2. Methods ST characteristics determined using the Cyclone Phase Space (CPS; Hart, 2003) diagnostics of ERA-40 and GFS analyses are a lower-tropospheric warm core and an upper-tropospheric cold core (Fig. 2). The initiation time of a ST is defined as the onset of gales, since weaker systems do not generate the societal peril and pub- Fig. 1. GOES-8 visible imagery and satellite-derived winds lic awareness that initially motivated this study. valid for 12:45 UTC October 12th 2001. Image courtesy of In this section, the datasets and model fields used Univ. Wisconsin CIMSS days confirmed that the Storm Warning issued did not capture the attention of the public as well as a Tropical Storm Warning likely would have. In the historical record of warm season storms affecting Bermuda, it is apparent that the nearby development of Karen was not a unique occurrence, and that it would be to the benefit of Bermudian forecasters to have a record of similar events. The purpose of this study is to examine all such ‘‘Bermuda storms’’ with subtropical characteristics in the last 49 years (for which there is reliable local observational data). Such subtropical cyclones (STs) may have been mistaken for tropical or extratropical systems, or poorly forecast. Subtropical storms are hybrid cyclones with cold upper and warm lower cyclonic components, spawned as baroclinic developments in the presence of positive low-level vorticity over relatively warm sea surface temperatures (SST). Guishard (2006) presents an Atlantic basin climatology of warm season cyclones with subtropical storm characteristics. This climatology was compiled through the initial detection and tracking of the cyclones in the 45-year ECMWF Reanalyses (ERA-40; Uppala et al, 2004) and corroborated with other evidence, where available. Fig. 2. Lifecycle evolution in the CPS of (a) a typical ST Using the methods described in Guishard which transitions into a TC and then undergoes extratropi- (2006), the characteristics of STs which have im- cal transition; and (b) an extratropical system which devel- pacted Bermuda are documented in this study. The ops tropical characteristics (tropical transition) and then objectives of this project are to (a) catalogue sub- undergoes extratropical transition. Note that while the position within the phase space is similar for each evolution, tropical systems which affected Bermuda with the direction of the path determines what the phase of the direct hits or ‘‘near misses’’; (b) elucidate any cyclone. A snapshot of the cyclone’s structure is not suffi- common characteristics in origin and evolution, cient to discern its subsequent evolution Bermuda subtropical storms to generate these analyses are reviewed and the must pass within 100 nm (185 km) of Bermuda. process of identification of STs from the general This distance corresponds to the operational list of warm season cyclones in the vicinity of ‘‘threat’’ criterion for the Bermuda marine area. Bermuda is elucidated. It is reasonable to suppose that a system within 100 nm could conceivably affect Bermuda with damaging winds and seas, change direction or in- 2.1 Datasets tensify quickly ‘catching people off guard’ (e.g., The compilation of storms for this study utilizes hurricane Emily in 1987; Gerrish, 1987). a range of datasets. Guishard (2006) describes an Gale force winds must be associated with Atlantic basin climatology based on 45 years of each candidate storm at some time during the the ERA-40 reanalyses (Uppala et al, 2004) and subtropical part of its lifecycle. Determination a set of rules developed to distinguish ST from of the presence or absence of gales is based on other warm season storm types. This coarser res- Bermuda and satellite observations of winds and olution study was complimented by a survey of pressure, and ERA-40=GFS gale algorithms). The the GFS operational analyses (also described in occurrence of gale force winds is operationally Guishard, 2006) for a five year period. The agree- considered the threshold for societal impacts due ment between these two methods of STs detec- to loss of life, injury, or property damage (e.g., tion in the overlapping time period, as well as Higgs, 2005). their concurrence with the recent NHC records There is disagreement between meteorologists (e.g., Stewart, 2000; 2001), provides confidence on the nature of ST genesis. Even hurricane spe- that the techniques are robust. These ERA-40 and cialists within NHC exhibit different operational GFS surveys are drawn upon for this study. opinions; one specialist’s ST storm is another’s Corroborating evidence of significant storm im- non-tropical low or extratropical system (personal pacts (if any) is then sought from historical ob- communications, Beven, Franklin, Roth, Stewart, servations over Bermuda, using the occurrence Landsea, Fogarty). For the purposes of this study, of gale force winds as a discriminator. Further, subtropical structural characteristics are distin- the Global ISCCP B1 Browse System (GIBBS) guished using CPS diagnostics of the global satellite dataset, available from the US National model analyses and satellite signatures. Satellite Climatic Data Center (Knapp, 2004), is used to signatures are determined using the methodology refine the determination of the overall synoptic of Hebert and Poteat (1975), who provide a ST characteristics (e.g., frontal, tropical, subtropi- analog to Dvorak (1975). Any period of purely cal). Storms which are ambiguous in structure extratropical or tropical structural characteristics (i.e., not clearly ST) and could be frontal in prior to the onset of gales for a candidate storm is nature have been discarded from the catalogue. limited to less than 24 hours. This ensures that the The current NHC Hurricane Database, HURDAT STs ultimately retained are not extratropically (NHC, 2006), was also examined to ensure the transitioning tropical cyclones (e.g., Evans and inclusion of all named storms which may have Hart, 2003) or decaying systems (either tropical had subtropical or ambiguous characteristics or extratropical). Extratropical gales are systems when they impacted Bermuda. The ST survey for which there is overall better documentation, compiled by David Roth of the Hydrometeoro- guidance and understanding amongst meteo- logical Prediction Center was also used (Roth, rologists, and as they are generally longer lived 2002) and a list of potential ST candidates for features, extratropical systems are more straight- inclusion in the reanalysis of HURDAT (Landsea forwardtoforecastthanST storms. If an ex- et al, 2003), recently compiled by Jack Beven of tratropical system with gale force winds gains the US National Hurricane Center, was also valu- tropical characteristics, it is deemed to have un- able for cross-checking candidate storms. dergone tropical transition (TT; Davis and Bosart, 2004). Likewise, a tropical cyclone (TC) which begins to exhibit elements of becoming extratro- 2.2 Identification of storms pical (even to the extent that it is at some point Criteria for inclusion of disturbances into this a hybrid system), it is deemed to be undergoing dataset are outlined in this section. The system extratropical transition (ET; Hart and Evans, M. P. Guishard et al

2001). The location of a system in the CPS at its which met the above criteria of being STs which initiation time (i.e., the onset of gales) and its affected the Bermuda area. In addition, 12 un- subsequent direction of movement through the named storms were found in the ERA-40 ST cli- CPS are both used to filter non-STs out of the matology of Guishard (2006) which matched the database. aforementioned discriminators, giving a total of 28 STs having affected Bermuda. The name of 3. Results each storm, date of closest point of approach to Bermuda, and the presence=absence of gale force 3.1 Spatial and temporal characteristics winds observed at Bermuda are presented in Table 1. There were 16 named storms already in the Inspection of the 28 ST cases identified indi- HURDAT Atlantic Hurricane Best Track dataset cates that the most favourable time of year for these systems is the height of the Atlantic hur- Table 1. List of the STs that affected Bermuda in the ricane season. Nearly half of the storms (13=28) 49 year period 1957–2005. Storms recorded in HURDAT as referred to in the text as ‘‘named’’ and storms newly represented here reached their closest point of identified in Guishard (2006) are referred to as ‘‘unnamed.’’ approach to Bermuda in September, climato- Symbols associated with unnamed storm numbers represent logically the peak of the Atlantic tropical sea- the surface wind speed recorded at Bermuda: Ã indicates son (e.g., Hart and Evans, 2001) and fully À1 $ À1 ? 17 m s , indicates 10–17 m s , and indicates no wind one quarter of the storms (7=28) formed dur- data available at Bermuda ing October. Three pre- and post-season storms Date Month Year Closest (Unnamed #1631352, December 1994; Unnamed approach to #1639403, May 2001; Ana, April 2003, Beven, Bermuda (km) 2003) are also present in this list, indicating that Storms recorded in HURDAT (‘‘named’’) the synoptic scale factors which dictate Atlantic Noname #9 16 October 1970 14 ST activity are not strictly confined to the hurri- Ginger 23 September 1971 107 cane season. Of course, the 2005 season also Alice 4 July 1973 24 demonstrated that TCs do not always obey the Dorothy 27 September 1977 64 season time limits either! Clara 10 September 1977 73 The average position for the onset of gales in Emily 2 September 1981 32 Debby 16 September 1982 114 the Bermuda STs identified in the ERA-40 is to Gustav 17 September 1984 14 the south through southwest of Bermuda; these Arlene 14 August 1987 50 storms generally then advance in a northeastward Grace 29 October 1991 50 direction, moving near the Island with a closest Karl 23 September 1998 64 point of approach to the northwest (Fig. 3). This Florence 16 September 2000 71 track is consistent with a steering flow dominated Karen 12 October 2001 12 Ana 18 April 2003 101 Nicole 11 October 2004 59 Harvey 4 August 2005 24 Subtropical storms identified in Guishard (2006) (‘‘unnamed’’) 16642 8 September 1959 162 16754 25 September 1960 101 164414 28 September 1963 165 166904Ã 30 September 1965 102 167735$ 30 June 1966 165 162971? 21 June 1972 170 1624795$ 25 October 1989 108 1627135Ã 5 September 1991 180 1631122Ã 16 October 1994 101 1631163$ 26 October 1994 64 $ 1631352 15 December 1994 101 1639403Ã 9 May 2001 64 Fig. 3. Tracks of the subtropical storms moving within 1 of Bermuda 1957–2002. Based on ERA-40 data Bermuda subtropical storms by the Bermuda High, the major subtropical presence of a name may mean that there is more steering feature in September and October. The availability of data such as aircraft reconnais- average position of origin in the ERA-40 clima- sance, additional satellite data and imagery, and tology is 31 N60 W (Guishard, 2006), which is CPS information from multiple models. actually to the southeast of the Island (but within In the next section, synoptic histories of one one model gridpoint), however, with this track named ST and one unnamed ST are discussed in pattern, storms forming to the east are likely con- more detail to give a context to the 49-year data- tinue to move away further east of the area. It is base presented here. These storms are typical of also notable that this position is only 470 km the 28 storm set. (255 nm) from Bermuda. Generally, storms named by NHC were per- ceived to have tropical, rather than subtropical, 3.2 Synoptic histories characteristics at the time of naming, although The number of STs found in this survey are too STs have been named occasionally in the NHC numerous to provide complete synoptic histories records (NHC, 2006). Indeed, the methods used at in this paper. Thus, as indicated above, the syn- NHC for the detection and naming of ST storms optic evolutions of two representative storms are are often subject to the individual interpretation provided: one represents a typical named storm of the situation of the forecaster on duty, where- that had subtropical characteristics in the CPS as the Guishard (2006) approach to determin- (Grace, October 1991). The other is unnamed ing the nature of cyclones is at least consistent. (#1639403 in Table 1) and thus is not present It is not the intent of the authors in this project in the HURDAT Best Track dataset. to criticize the diligent and careful work of the NHC hurricane specialists; certainly the decisions (1) Hurricane Grace, October 1991 made in an operational forecast environment On October 25th, a mid-level low developed may be at odds with a post-event analysis. Our down to the surface, forming a subtropical de- goal is to use the benefit of hindsight to provide pression 578 km south of Bermuda at 18 UTC forecasters in Bermuda with a more complete pic- (Rappaport, 1991). This depression developed ture of the STs likely to affect their region. quickly into hurricane Grace, with winds attain- Out of the total 28 STs in this study, 16 were at ing hurricane force by the time the storm reached some point named by NHC (and so are referred 32 N, at 00 UTC on the 28th, just 330 km west to as the ‘‘named’’ subset). In the remainder of of Bermuda. Shortly after becoming a hurricane, this paper, the storms found here will be treated Grace was quickly swept eastward by a cold as being in the same category whether named or front, passing 78 km to the south of Bermuda unnamed, but names will be used wherever pos- as a Category-1 hurricane. Local observations sible, for ease of recognition. Nonetheless, the on the 29th recorded a peak mean wind speed

Fig. 4. Two views of hurricane Grace (1991): (a) NHC Best Track (image courtesy of weather.unisys.com); and (b) GOES IR satellite image valid at 00 UTC on October 29th 1991 (source: US National Climatic Data Center) M. P. Guishard et al of 21 m sÀ1 with a peak gust of 29 m sÀ1. A storm centre was exposed on the 26th, indicative of the total of 86.61 mm of rainfall was collected at the storm’s subtropical nature at this time. Central airport from October 26th through 29th. convection then began to grow as Grace’s tropi- The Best Track of Grace along with an infra- cal characteristics increased, followed by a rain- red satellite image of Grace valid at 06 UTC on band forming at 12 UTC on the 27th. Cyclonic October 29th, 1991 are depicted in Fig. 4. winds decreasing with height were evident on the At the onset of gales (00 UTC on the 26th), radiosonde ascent at 12 UTC on the 29th, indi- ERA-40 analyses indicate that the surface tem- cating that the storm had become tropical by the perature was 26 C and the 500 hPa temperature time it affected Bermuda. was À8 C. This difference of 34 C indicates Grace went on to contribute to damaging a reasonable amount of instability in the lower storms in the New England and Canadian Mar- levels. A deep, sharp, north–south oriented upper itimes region and could certainly have been a level trough lay 3 to the west, with an isohypse greater threat to Bermuda, had the timing of of 960 dam at 300 hPa and a latitudinal extent the larger-scale upper trough which changed the of 10, producing southwesterly shear over the storm’s course been a few hours slower. This region of ST formation. The trough was cut off would have allowed the storm to continue along at 500 hPa. A cold core was apparent down to its earlier northerly track for a few more miles, 500 hPa, and a warm core up to 850 hPa. The and could possibly have led to hurricane force 300 hPa trough became a cut-off upper low at winds directly over the Bermuda area. This sce- 00 UTC on the 27th. At this point the track chan- nario would have also delayed the onset of shear- ged from northward to eastward, and ST Grace ing from the upper trough, which could have now approached Bermuda from the west. The sys- allowed for additional strengthening of the storm. tem became warm core to 300 hPa by 06 UTC (2) Unnamed subtropical storm May 2001 on the 28th, until which time areas of cold and The track (derived from ERA-40) and associated warm advection could be seen in the 1000– GOES IR satellite imagery for the unnamed ST 500 hPa thickness pattern to the west and east outlined in this subsection are displayed in Fig. 5. of the storm respectively. By 18 UTC on the A closed low is ﬁrst detected in the ERA-40 28th this cut-off upper low was picked up by ﬁelds at 12 UTC on the 5th, and the presence of a larger-scale upper trough and moved out of gales by 18 UTC on the same day. At this time, the region. At this point the track changed from the re-analyses represent that the surface tem- northerly to easterly, which led ST Grace to ap- perature was 22 C, and the 500 hPa temperature proach Bermuda from the west. was À10 C, a difference of 32 C. A 948 dam In the visible loop, the trough showed few northwest–southeast oriented trough, with a hor- signs of organization on the 25th. The low-level izontal extent of 10, lay 3 west-northwest of

Fig. 5. The unnamed STof May 2001: (a) ERA-40 track of the unnamed ST from May 5–18 2001 (the location of Bermuda is highlighted with an X); and (b) GOES IR satellite imagery valid at 12 UTC on May 9th 2001 (source: US National Climatic Data Center) Bermuda subtropical storms the center of the surface low, which was located 700 hPa, and increased to 26 m sÀ1 at 300 hPa, at 26 N, 69 W at this time. confirming the subtropical nature of the storm. The upper trough continued to dig further The system underwent reorganization on the south, generating a cut-off upper low on the 6th. 9th, as the low immediately to the south of the The upper low moved around Bermuda to the Island deepened to become the main feature, and south, and was to the southeast of the Island at continued to move towards Bermuda, passing 06 UTC on the 7th, after which it was absorbed by very close to the west of the Island. This low then an upper trough moving by to the north. This next began to occlude as the storm moved away to the trough then strengthened and generated another north. Storm total precipitation was 43.4 mm, cut-off upper low (936 dam) by 06 UTC on the mostly on the 8th, when widespread thunder- 8th. This last cut-off low also moved south of the storms were observed overnight. Island on the 9th, finally being swept north by ERA-40 sea level pressure and surface wind another deep trough to the west of the Island early charts show a large and asymmetric radius of on the 10th. gale force winds throughout the storm’s life cy- Inspection of surface analysis charts reveals that cle. This surface structure would typically be asso- an occluded surface low that lay to south began to ciated with an extratropical feature. The storm was approach the Bermuda area on May 7th. On the cold-core until 18 UTC on the 9th, when a low- 8th, a second surface low developed to the north level warm core can be seen in the surface charts of the original low, at the point where the occlu- and in the CPS analyses (Guishard, 2006). The sion met the cold and warm fronts (Fig. 6). The thickness charts depict a clearly baroclinic sys- formation of the second low, which lay immedi- tem initially, however, there was a weak tempera- ately to the south of the Island, created a high pres- ture gradient on the 7th. Cold air dug down again sure gradient on its northern side. Therefore, the on the western side of the low on the 8th, remain- highest wind speeds were recorded at this time. A ing north of the center of the low on the 9th. mean wind speed of 16 m sÀ1 from the east-north- In the visible images from the 9th, the low- east, with a gust to 22 m sÀ1, was recorded at the level circulation is apparent. There was a small airport at 13 UTC. At North Rock, an offshore area of central convection just northeast of the wind sensor, a gust to 28 m sÀ1 was also recorded. centre of the surface low, and the cold front was The 12 UTC Bermuda radiosonde ascent on the removed by about 3 from the centre of the low 8th recorded winds that decreased to 10 m sÀ1 at (Fig. 5). The storm became increasingly extratropical as it moved northwards. As the storm passed by the Island, the ST satellite signature was analyzed as a Hebert-Poteat (HP) estimate of 3, corresponding to estimated surface winds of 23–26 m sÀ1. To confirm the actual subtropical nature of this storm we must refer to observational evidence. Satellite images, synoptic patterns, and the Bermuda ascent depict both tropical and extratropical characteristics. This evidence is further supported by the storm’s trajectory through the CPS (not shown). Therefore, this storm is another example of a ST that was a threat to Bermuda, and could have produced greater damage if it had passed directly over the Island.

3.3 Bermuda STs in the context of Atlantic ST climatology

Fig. 6. The unnamed ST of May 2001: NCEP surface ana- The subset of Atlantic STs affecting Bermuda lysis valid at 00 UTC on May 8th 2001 and the complete set of Atlantic STs documented M. P. Guishard et al in Guishard (2006) are compared here to focus tation of the trough is slightly positive, as dis- on the most likely behaviour of Bermuda STsas cussed in Sect. 5. As noted in the introduction, a guide for local forecasting purposes. subtropical cyclones are hybrid systems with Guishard (2006) found that the orientation of cold-cored upper- and warm-cored lower-tro- the upper trough associated with the incipient pospheric thermal signatures. These storms are near-surface vortex did not correlate well with spawned as baroclinic developments in the pre- the thermal structure – only that an upper trough sence of positive low-level vorticity over rela- or cyclone was present during the development tively warm SST or strong SST gradients. The of all ST storms. Specifically, in the 1998–2002 evolution of a typical ST is illustrated in the sche- GFS survey, trough axes varied from strongly matic in Fig. 7. The initial baroclinic cyclogen- positive to neutral to strongly negative in hori- esis is illustrated in panel (a) and the completion zontal orientation. The troughs all had at most a of transition to a subtropical cyclone is depicted 960 dam 300 hPa height (the lowest, Karen, hav- in panel (b). In (b), the upper trough has cut off ing its lowest kinked isohypse of 948 dam). In and the surface occlusion (with its associated this study, however, the average horizontal orien- cloud shield) has become detached from the main baroclinic zone. At this time, the system is deemed to have become a subtropical cyclone: the near-surface feature has become distinctly more vertically-stacked in nature, and the upper feature has cut off from the westerlies as a cold low, indicated by the X. In the process of the upper cyclone becoming cut-off, it is reduced in scale from macro--scale to meso--scale in horizontal extent (that is, from a synoptic to a hurricane-scale low) (see e.g., Orlanski, 1975 or Thunis and Bornstein, 1996). At the same time, the diameter of the surface low has increased from a meso--scale feature to meso--scale. This scale matching is necessary for the ST development, yet remains poorly understood. The wind structure of a ST supports the idea of a hybrid cyclone, through thermal wind arguments. For example, in the ascent data taken during Karen (2001) (see Guishard, 2006, and Williams, 2002), there was a near-surface cyclone of 30 m sÀ1 that decreased with height to approximately 700 hPa, and then increased again to 20 m sÀ1 at 300 hPa. This hybrid structure remains evident until one of three conditions is satisfied: the convection erodes the upper vorticity maximum, the convection is suppressed, or another trough interacts with the evolving system. Fig. 7. Schematic of the synoptic evolution lifecycle of an If the convection is able to erodes the upper Atlantic subtropical cyclone for (a) the initial baroclinic vorticity maximum to the point when the lower cyclogenesis, and (b) the evolution to a subtropical cyclone. warm core dominates the system, low-level con- Solid black lines are surface isobars, with an L at the posi- vergence and deep convection begin to take off, tion of the central low pressure. Solid arrows are upper and the storm becomes more tropical in nature streamlines, for example, the 300 hPa flow. Surface fronts are marked in the conventional manner, and hatching indi- (e.g., Davis and Bosart, 2003). Ultimately, this cates the continuous cirrus shield associated with ascent. may lead to a tropical cyclone, with a symmetric The dashed line is the upper trough axis warm core extending upwards through the sys- Bermuda subtropical storms tem’s depth, an upper anticyclonic outflow, and lower circulation that is the case in the tropical convection which has completely wrapped around scenario just described. If this cold low remains the central low. While forecasters have been aware cut off, it will fill in a matter of days. Thus, loss of this mechanism of tropical cyclogenesis for of convection without another external forcing some time, it is only recently that the frequency mechanism results in the ultimate dissipation of of occurrence of such genesis has been documen- the ST. ted for the North Atlantic STs (Guishard, 2006). The two lifecycles described above can only Should the convection not be sustained (e.g., occur in the absence of another upper trough ap- due to cold SST, or dry intrusions), the cyclone proaching. In the event that the cyclone becomes may become more extratropical in character. If mobile under the influence of the steering flow of this is the case, the upper cold low may build a new trough, it will quickly transition to a fully down to the surface, leading to a less intense extratropical low.

Table 2. Necessary but not sufﬁcient conditions for subtropical versus tropical cyclogenesis

Subtropical cyclone (ST) Tropical cyclone (TC)

Dynamic conditions Upper trough (non-trivial deep layer shear) Weak deep layer shear Anomalous, positive low level relative vorticity Anomalous, positive low level relative vorticity Latitudes 20–35 N (Coriolis large cf with TC genesis) Latitudes 5–15 N Thermodynamic conditions Deep convection possible with forced ascent. Due to warm SST Ability to sustain deep convection facilitated and cold upper temperatures, near-neutral stability in free troposphere by warm SST, saturation in the boundary layer Lower troposphere moisture anomalously high cf long-term Lower troposphere moisture anomalously high monthly average cf long-term monthly average

Table 3. Selected GFS analysis parameters associated with 18 Atlantic ST during the period 1999–2004 (Guishard, 2006). Storms labeled with an asterisk (Ã) do not appear in the NHC Best Track Database, HURDAT

Storm 1st occurrence of gales Position SST Shear B Vtl Vtu (C) (m sÀ1) N W

1 Ana 12 UTC 17 Apr 2003 27.3 68.4 23.7 19.2 15.9 À3.1 À103.3 2 Cristobal 00 UTC 7 Aug 2002 30.0 76.0 29.0 9.2 0.5 79.3 À43.0 3 Florence 12 UTC 12 Sep 2000 31.0 73.0 27.9 8.7 À10.1 32.5 10.1 4 Gabrielle 00 UTC 14 Sep 2001 24.9 85.0 29.7 6.8 19.1 48.5 À40.8 5 Gustav 12 UTC 8 Sep 2002 27.9 70.1 29.3 18.2 7.8 49.9 À52.6 6 Juan 00 UTC 26 Sep 2003 30.5 61.1 27.9 9.9 8.8 36.3 À9.1 7 Karen 12 UTC 11 Oct 2001 29.2 62.6 28.2 21.2 28.2 36.0 À218.6 8 Kyle 00 UTC 22 Sep 2002 31.9 51.0 26.9 6.5 À0.5 45.1 27.7 9 Leslie 12 UTC 5 Oct 2000 29.9 76.9 27.8 5.2 6.4 46.8 À26.6 10 Lorenzo 12 UTC 28 Oct 2001 27.1 39.7 26.6 13.7 1.2 16.2 À74.5 11 Michael 12 UTC 13 Oct 2000 26.1 69.9 28.8 28.8 22.9 49.0 À216.6 12 Nicole 12 UTC 9 Oct 2004 28.3 62.9 28.0 17.2 À0.1 32.2 À140.6 13 Noel 00 UTC 3 Nov 2001 32.9 43.7 25.22 8.5 6.4 54.6 À158.1 14 Olga 12 UTC 23 Nov 2001 30.0 50.0 24.4 8.1 15.1 41.6 À69.1 15 Otto 00 UTC 26 Nov 2004 29.4 37.9 24.3 20.8 À8.9 À72.6 À145.4 16 Peter 00 UTC 06 Dec 2003 29.7 36.8 23.3 16.9 12.4 À8.6 À117.6 17 ST August 2000 12 UTC 29 Aug 2000Ã 28.5 77.1 29.0 11.4 11.8 69.8 À58.4 18 ST October 2000 00 UTC 1 Oct 2000Ã 27.0 77.0 29.1 19.9 11.9 À16.1 À78.4 Average 28 Sep 29.0 62.2 27.2 13.9 8.3 29.8 À84.2 Std. Deviation 51.82 2.04 15.3 2.1 6.6 10.2 35.9 70.9 M. P. Guishard et al

4. Subtropical vs. tropical storms that have affected Bermuda Despite the differences in structure between TCs and STs there are similarities in their societal effects and their potentially erratic movement. The significant weather produced by each storm is the same, including cumulonimbi, gale force winds, and heavy rainfall – although the magni- tudes of these effects may differ. In this section, comparisons are made between TCs and STs, with a particular focus on those systems which have affected Bermuda. The necessary (but not sufficient) conditions Fig. 8. NHC forecast track errors for hurricanes Fabian for the ST formation are contrasted with the (2003), Nate (2005), and the 1995–2004 decadal average conditions needed for tropical cyclogenesis in for 12–120 hour forecast lead times (data from Pasch et al, Table 2. Note that the main difference in the cri- 2004, and Stewart, 2005) teria for each type of disturbance is the necessity for shear in ST genesis, as opposed to the need with the track errors for Nate (2005), a storm for weak shear for TC genesis and development of subtropical origin (Stewart, 2005) are depic- (Gray, 1968; McBride and Zehr, 1981) and the ted in Fig. 8. The excellent track forecasts for forcing necessary to sustain deep convection (see Fabian (2003) maximized the preparation time also Guishard, 2006). for residents of Bermuda, reducing the danger Tropical cyclones that have affected Bermuda to Bermuda from the storm. The problematic range in intensity from weak tropical storm to forecast of Nate (2005) illustrates that such track strong Category-4 (September 21st, 1922) on the forecast skill is not the rule for STs (Fig. 8). Saffir-Simpson scale. Hurricane Fabian struck Indeed, the forecast track of Fabian is also con- Bermuda on September 5th, 2003 with wind trasted with that of Karen, whose track was not gusts recorded to 76 m sÀ1 and storm surge to forecast until two days after the formation (ac- 3.4 m above high tide, resulting in the loss of four cording to the NHC advisories) in Fig. 9. Often, lives and widespread damage. Fabian was a clas- TCs which are as far north as Bermuda (such as sical Cape Verde hurricane which intensified to Fabian) are already under a strong steering flow, Category-4 on the Saffir-Simpson scale and re- and are therefore more predictable in terms of curved over Bermuda as a Category-3 hurricane motion. Contrast this with the synoptic blocking (Pasch et al, 2004). pattern necessary for subtropical cyclogenesis Track errors for hurricane Fabian (2003) versus (see previous section) and the difficulty of ST the average track errors for 1993–2002, along (such as Karen) track forecasts becomes more

Fig. 9. Post-storm ‘‘best track’’ and NHC forecast track for (left) hurricane Fabian (2003) and (right) hurricane Karen (2001) Bermuda subtropical storms understandable. The weak steering flow typical Spring storms appear more likely to have sur- of ST genesis and the lack of forecaster aware- face precursors associated with shear vorticity ness of this fact has historically made the STs zones along old fronts. An occlusion (like that which affect Bermuda less predictable than TCs; during Karen’s onset; Hulme and Martin, 2006) this is especially true in light of recent improve- or a preexisting area of convection (e.g., a con- ments in TC track forecasting. vergence line) often interact with an upper trough to the west, reducing the shear over the develop- 5. Summary and discussion ing surface center. Lower-mid tropospheric warm-core intensifi- September and October are the peak months of cation necessary to attaining the hybrid structure occurrence of STs near Bermuda (Fig. 10), which of a ST (rather than a purely extratropical sys- is similar to that of the North Atlantic basin as tem) relies on sustained convection (Guishard, a whole. The temporal distribution of ST storms 2006). This may be driven by a combination of affecting Bermuda is centered on September, warm SST and forced ascent (e.g., ahead of the with half the storms occurring in that month. trough). The Gulf Stream was seen, in the case of Another quarter of storms affect Bermuda in #1631122, to provide the warm surface water October. Storms otherwise have occurred outside required to enhance the lower level warm core. the season in April and May and post-season in STs were noted, in ERA-40 upper air charts December. In the last 49 years, the only hurri- to develop under the diffluent eastern side of a cane season month in which Bermuda has not NNE through SSW oriented shortwave upper been affected is November. trough, which is cut off in the cases of most STs which threaten Bermuda often form just rapid development. The upper trough orientation southwest of the Island, deepening quickly into is generally south-southwest to north-northeast, hurricanes by the time they reach the Island (e.g., located 2 to 4 degrees west-northwest. The inci- Noname #9 1970 and Alice 1973). The average pient surface cyclone develops warm core char- position for the onset of gales from STs affecting acteristics over warm waters, due to convection. Bermuda is to the south through southwest of the This mechanism seems to occur whether asso- Island. The storms generally then advance in a ciated with a cut-off low or shortwave trough northeastward direction, moving past the Island in the upper levels. Composite 300 hPa heights with a closest point of approach (or CPA) to the and mean sea level pressures from 18 ST storms northwest (Fig. 3). This track is consistent with (Table 3) documented by Guishard (2006) reveal- steering dominated by the climatological position ing that an upper cyclonic anomaly of À6 dam on and extent of the Bermuda High in the months of average is present at 300 hPa, at the onset of September and October. In contrast, the average gales (Figs. 11 and 12). In the Bermuda case position of origin for all Atlantic ST is 31 N studies, the disturbances are associated with 60 W (southeast of Bermuda by 470 km or 300 hPa troughs which invigorate surface lows 250 nm) (Guishard, 2006). that develop as ST storms. Most frequently, they are picked up by a larger scale upper trough and pulled north (Noname #9, Karen) towards the

Unnamed Island. However, if this fails to occur, they may Named end up drifting eastwards (e.g., TS Arlene, 1987). In the initiation stage, the movement of the upper feature must phase with that of the surface disturbance for ST development to occur. Moderate deep layer vertical shear must also exist, which has been indicated as a necessary condition of tropical transition development in previous studies (Davis and Bosart 2003). Fig. 10. Total numbers of STs by month for 1957–2005. The average SST for the storms in this study Named (unnamed) storms are indicated by the white (black) is 26 C. Based on the coldest ERA-40 500 hPa bars temperature, the difference between the SST and M. P. Guishard et al

Fig. 11. Composites of the GFS analyses 300 hPa geopotential heights and sea level pressures of 18 STs in Table 3 at: (a)24 hours prior to the onset of gale-force winds; (b) the ﬁrst occurrence of gale-force winds; and (c) 24 hour after the onset of gale-force winds. The 300 hPa geopotential height anomalies and sea level pressure anomalies for the same times are plotted in panels (d), (e), and (f), respectively

Fig. 12. Composites of vertical cross-sections of potential vorticity and potential temperature (a) 24 hours prior to the onset of gale-force winds; (b) at the onset of gale-force winds; and (c) 24 hours after the onset of gale-force winds. Composites are of the 18 STs in Table 3. The dynamic tropopause is represented by the 2 PVU contour (darkest shading) the 500 hPa temperature is 34 C for the Bermuda arguments), over a relatively warm SST. The SST ST cases. Decreases in static stability have also distribution for ST formation in the ERA-40 cli- been identiﬁed by Guishard (2006) as common to matology is skewed towards higher SST (Fig. 13) the ERA-40 ST climatology. This is unsurprising but note that the occasional ST forms over colder due to the movement of an upper cyclonic anom- waters (Fig. 14). Also, hybrid cyclones have been aly (necessarily a cold feature, from thermal wind documented over colder bodies of water than Bermuda subtropical storms

Fig. 13. Frequency distribution of SST, for all ERA-40 STs (Source: Guishard, 2006)

Fig. 14. Flow chart of descriptors based on synoptic analyses of each of the 28 storms listed in Table 1. HP class numbers (representing the following wind ranges) are printed in parentheses on the chart: 2.5: 18–21 m sÀ1; 3.0: 23–26 m sÀ1; 3.5: 28– 34 m sÀ1. In addition, the total numbers and percentages of the ERA-40 climatology in each of the four categories are listed the subtropical Atlantic, e.g., ‘‘hurricane Huron’’ Only a few Bermuda STs reached peak in- (Miner et al, 2000), Mediterranean storms, South tensity beyond gale force in the vicinity of the Atlantic storms, such as Catarina in March 2004 Island; Alice, 9, and Grace are the named storms (McTaggart-Cowan et al, 2005). In these situ- which passed near Bermuda with winds of hurri- ations, the column is rendered less stable by an cane force, all of them attaining Category 1 on upper cold feature. Guishard (2006) also pro- the Saffir-Simpson scale. The location of genesis posed that, in the absence of tropically warm is typically 3 to the southeast through southwest SST, strong SST gradients would provide warm of Bermuda and after genesis most STs move and cold advection around the developing sys- north past the island. The distance between the tem. This was also observed in the South Atlantic average position of the onset of gales and the storm, Catarina (McTaggart-Cowan et al, 2005). average storm closest position of approach to M. P. Guishard et al the Island is on the order of half a degree or less. 1957, the integrity of cyclone resolution in the The frequency of gales (>17 m sÀ1) from STsat ERA-40 sub- and extratropics is sound compared Bermuda is 4 in 50 years (i.e., slightly less than to data for the tropics and polar regions. The one per decade). Those four are Alice (1973), named storms do not overlap with the automated Karen (2001), Nicole (2004), and Unnamed storm tracks closely enough to confirm the effec- ERA-40 storm #1631122 (1994); of these, tiveness of the ERA-40 tracking in isolation. Karen was the only ST storm to have produced Thus, confirmation of ST effects through surface more than 25 m sÀ1 at Bermuda. wind records and, where available, radiosondes The greatest threats to Bermuda were posed by from Bermuda, is a necessary component of Noname #9 (1970), Alice (1973), Grace (1991), this study. and Karen (2001). Karen caused the most damage Subtropical storms are as much of a threat due to its quick development close by. The stron- regarding wind strengths in the tropical storm gest winds experienced from the STs identi- range as TCs. Lack of predictability makes them fied from the ERA-40 (over 15 m sÀ1 recorded as dangerous as tropical storms, and they occur at at Bermuda) were from storms 1631122 (16 the same time of year; sometimes the focus can October 1994) and 1639403 (9 May 2001), both be more on classically tropical systems. Forecas- of which were fairly strong storms (HP 345– ters must not be complacent about disturbances 50 23–26 m sÀ1) whose centres passed within originating from upper cyclones in the subtrop- 100 km and 64 km of Bermuda, respectively. ics, especially during Hurricane Season. Vigi- While the 2001 storm formed in a typical loca- lance for these systems must even be maintained tion for ST storms, to the south of the Island, the outside of hurricane season, as illustrated by 1994 storm formed over the Gulf Stream, in occasional ST occurrence in every other month the region typical for Hatteras low formation (Roth, 2002; Guishard, 2006). Thanks to the (Vandever, 1994). Storm 1627928 (21 May 1992) NHC, BWS, and the WMO, awareness of such produced a high recorded windspeed, 15 m sÀ1, systems has been raised in recent years; STs are with a gust of 20 m sÀ1, while still quite far away, now named from the same list as TCs in the and passing to the east, and has the highest HP Atlantic, and receive increased observations and estimate for maximum storm wind speed. This analysis as a result. storm’s slow forward speed could help to account for its rapid formation. This rapid development and the May occurrence both added to the sur- Acknowledgements prise factor, making this a difficult forecast. The authors benefited from lively discussions with Jessica While the effects of none of these storms rival Higgs (Arnoldy), Justin Arnott, and Adam Moyer, as well as those of Karen, some of them would have had valuable input from Jack Beven and David Roth. Charles Pavloski’s aid with model data management was of great similar effects had their track taken them some- help. Many analyses and figures were created using the Grid what closer to Bermuda, according to HP esti- Analysis and Display System (GrADS). mates, along with wind speed recorded at the Financial support was granted for the first author’s research Island compared to the CPA, and speed of for- by the Government of Bermuda Ministry of Transport and mation apparent in satellite photos. The Karen Tourism (Department of Airport Operations). BAS-Serco Ltd. also helped to fund the contribution of the authors on its staff. phenomenon, a strong subtropical storm passing The third author’s participation in this work was supported very close by Bermuda, and resulting in recorded by the National Science Foundation under grant no. ATM- wind speeds over 26 m sÀ1, stands out in this 49 0351926. This financial assistance is greatly appreciated. year climatology. The ERA-40 is subject to large errors in intensity and position (Bengtsson et al, 2004), parti- References cularly over data-sparse areas, like the ocean, and Bengtsson L, Hodges KI, Hagemann S (2004) Sensitivity particularly in the case of a storm over the ocean, of the ERA-40 re-analysis to the observing system: de- which marine vessels would purposely steer clear termination of the global atmospheric circulation from reduced observations. Tellus 56A: 456–471 of. Nonetheless, Bengtsson et al (2004) indicate Beven J (2003) Tropical storm Ana, 20–24 April 2003. NHC that because of the availability of reliable ra- Tropical Cyclone Report. Available at http:==www.nhc. diosonde data and aircraft reconnaissance from noaa.gov=2003ana.shtml Bermuda subtropical storms

Davis CA, Bosart LF (2003) Baroclinically induced tropical McTaggart-Cowan R, Bosart L, Davis C, Atallah E, Gyakum cyclogenesis. Mon Wea Rev 131: 2730–2747 J (2005) Analysis of hurricane Catarina (2004). North- Davis CA, Bosart LF (2004) The TT problem: forecasting east Tropical Cyclone Workshop, Rensselairville, NY, the tropical transition of cyclones. Bull Amer Meteor Soc June 2005. Available at http:==www.atmos.albany.edu= 85: 1657–1662 facstaff=rmctc=neHurrConf.pdf Dvorak VF (1975) Tropical cyclone intensity analysis and Miner T, Sousounis PJ, Wallman J, Mann G (2000) forecasting from satellite imagery. Mon Wea Rev 103: Hurricane Huron. Bull Amer Meteor Soc 81: 223–236 420–430 Moyer A (2006) Determination of observed gale radius Evans JL, Hart RE (2003) Objective indicators of the life- statistics for North Atlantic tropical cyclones. Master’s cycle evolution of extratropical transition for Atlantic Thesis, Department of Meteorology, The Pennsylvania tropical cyclones. Mon Wea Rev 131: 909–925 State University Gerrish HP (1987) Hurricane Emily, 20–26 September. NHC (2006) Hurricane best track files (HURDAT) NHC Preliminary Tropical Cyclone Report. Available 1851–2005. Available at http:==www.nhc.noaa.gov= from http:==www.nhc.noaa.gov=archive=storm_wallets= tracks1851to2005_atl.txt atlantic=atl1987-prelim= Orlanski I (1975) A rational subdivision of scales for atmo- Gray WM (1968) Global view of the origin of tropical spheric processes. Bull Amer Meteor Soc 56: 527–530 disturbances and storms. Mon Wea Rev 96: 669–700 Pasch RJ, Blake ES, Brown DP (2004) Hurricane Fabian, Guishard MP (2006) Atlantic subtropical storms: climatol- 27 August–8 September 2003. NHC Tropical Cyclone ogy and characteristics. Doctoral thesis, Department of Report. Available from http:==www.nhc.noaa.gov= Meteorology, The Pennsylvania State University 2003fabian.shtml Guishard MP, Evans JL, Hart RE (2006) Atlantic subtropical Rappaport E (1991) Hurricane Grace, 25–29 October 1991. storm climatology. J Climate (submitted) NHC Tropical Cyclone Report. Available from http:== Hart RE (2003) A cyclone phase space derived from thermal www.nhc.noaa.gov=archive=storm_wallets=atlantic= wind and thermal asymmetry. Mon Wea Rev 131: 585–616 atl1991-prelim=grace= Hart RE, Evans JL (2001) A climatology of the extratropical Roth DM (2002) A fifty year history of subtropical cy- transition of Atlantic tropical cyclones. J Climate 14: clones. 25th AMS Conf. on Hurricanes and Tropical 546–564 Meteorology, P1.43 Hebert PH, Poteat KO (1975) A satellite classification Stewart SR (2000) Hurricane Michael, 17–19 October technique for subtropical cyclones. NOAA Technical 2000. NHC Tropical Cyclone Report. Available from Memorandum NWS SR-83 www.nhc.noaa.gov Higgs JL (2005) A study of the evolution of the radius of Stewart SR (2001) Hurricane Karen, 12–15 October 2001. gale force winds using principle component analysis and NHC Tropical Cyclone Report. Available from http:== sliced inverse regression. Master’s Thesis, Department of www.nhc.noaa.gov=2001karen.html Meteorology, The Pennsylvania State University Stewart SR (2005) Hurricane Nate, 5–10 September 2005. Hulme AL, Martin JE (2006) The role of the occlu- NHC Tropical Cyclone Report. Available from http:== sion process in the extratropical-to-tropical transition of www.nhc.noaa.gov=2005atlan.shtml Atlantic hurricane Karen. 27th AMS Conf. on Hurricanes Thunis P, Bornstein R (1996) Hierarchy of mesoscale flow and Tropical Meteorology, Monterey, CA, 23–28 April assumptions and equations. J Atmos Sci 53: 380–397 2006, 10B.2 Uppala S, Kallberg P, Hernandez A, Saarinen S, Fiorino M, Knapp KR (2004) ISCCP B1 Data at NCDC: a new climate Li X, Onogi K, Sokka N, Andrae U, Bechtold VDC (2004) resource. 13th Satellite Meteorology and Oceanogra- ERA-40: ECMWF 45-year reanalysis of the global at- phy Conf., Norfolk, VA, Amer Meteor Soc mosphere and surface conditions 1957–2002. ECMWF Landsea CW, Anderson C, Charles N, Clark G, Dunion J, Newsletter 101: 2–21 Fernandez-Partagas J, Hungerford P, Neumann C, Vandever MR, Pearson DS (1994) Forecaster’s handbook for Zimmer M (2003) The Atlantic Hurricane Database Re- Bermuda. Naval Atlantic Meteorology and Oceanography analysis Project-Documentation for 1851–1910 altera- Facility, Bermuda, 101 p tions and additions to the HURDAT database. Hurricanes Williams R (2002) Bermuda Weather Service tropical cy- and Typhoons: Past, Present, and Future (Murnane RJ, clone report on hurricane Karen. WMO RAIV=HC-XXIV Liu KB, eds). Columbia University Press 2002 Doc 4.2 McBride JL, Zehr R (1981) Observational analysis of tropical cyclone formation. Part 2: Comparison of non- Corresponding author’s address: Mark P. Guishard, developing versus developing systems. J Atmos Sci 38: Bermuda Weather Service, BAS-Serco Ltd., St. Georges, 1132–1151 Bermuda (E-mail: [email protected])

View publication stats