Subtropical cyclones (ST) have only recently gained attention as damaging weather systems. A set of criteria for identifying and classifying these systems is introduced here and employed to identify 18 ST cases forming in the 1999–2004 hurricane seasons. To be classified as an ST, these systems must have near-surface gale-force winds and show hybrid structure for more than one diurnal cycle. The 18 ST cases are partitioned into four classes based upon their genesis environments. Genesis over waters with SST in excess of 25°C is observed in almost 80% of warm-season cases, in contrast with only 55% in an ST climatology presented in a companion study. The low-shear magnitude constraint recognized for tropical cyclogenesis is less apparent for ST formation with over 50% forming in the two partitions characterized by shear in excess of 10 m s−1. This relatively high-shear environment corresponds to equatorward intrusion of upper troughs over the relatively warm SST present in the mid–late hurricane season. Anomaly composites confirm that ST genesis is associated with the intrusion of an upper trough in the westerlies into a region of relatively warm SST and weak static stability, with a corresponding reduction in the environmental shear near the time of ST genesis. These conditions correspond well with the conditions for tropical transition identified by Davis and Bosart. Indeed, these systems exhibit a propensity to continue development into a tropical cyclone; 80% eventually became named tropical systems. This result is consistent with a recent ST climatology but had not been widely recognized previously. This raises the possibility that tropical storms evolving from ST may have been overlooked or their tracks truncated in the National Hurricane Center Hurricane Database (HURDAT).
Subtropical cyclones (ST) have a checkered history in the North Atlantic. They are intermittently recorded in the National Hurricane Center (NHC) Hurricane Database (HURDAT), but these records are incomplete (Guishard et al. 2007). Recent ST landfalls, such as Hurricane Karen (2001) in Bermuda, highlight the severe weather impacts associated with these events. Indeed, the media outcry after Karen’s landfall revealed that this storm was interpreted by the general public to be an unnamed tropical storm (Guishard et al. 2007).
In this study, we utilize a series of North Atlantic ST case studies to develop a uniform set of ST criteria that are well resolved using operational analyses. The structural characteristics and life cycle evolution of these storms are examined using Global Forecast System (GFS) operational analyses (Kanamitsu 1989) and National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis fields (for anomaly calculations; Kalnay et al. 1996), as well as a variety of satellite and other data sources. A conceptual model for ST formation based on a hybrid thermal structure in the cyclone phase space (CPS; Hart 2003) is introduced here. This conceptual model is applied to differentiate these case studies from other candidate storms. The ST criteria presented here were applied to the 40-yr European Centre for Medium-Range Weather Forecasts Re-Analysis (ERA-40; Uppala et al. 2004) in an accompanying study (Guishard et al. 2009, hereafter Part II) to derive a 45-yr climatology of North Atlantic ST for the period September 1957–August 2002.
Datasets used in this study and our analysis methodology are documented in section 2. Subtropical cyclone thermal and surface wind structures are described in section 3, and the criteria applied to identify ST are documented in section 4. The temporal and spatial characteristics of 18 ST cases are presented in section 5, and the results of storm-centered composites are reviewed in section 6. In section 7, potential paths to ST genesis are introduced, and a stratification of ST based upon their genesis environments is proposed and illustrated in section 8 using the 18 ST cases studied here. This environmental stratification is one step toward development of an operational forecasting strategy for these storms. Finally, these results are summarized in section 9.
2. Datasets used
A set of individual storm case studies is examined here with the objective of examining ST structure itself, along with the meso- to synoptic-scale environments necessary for such developments. The storms presented here meet the definition of ST (section 4; Part II), but they have developed in a continuum of formation environments (Guishard 2006).
a. Model analyses utilized
NCEP GFS (Kanamitsu 1989) gridded operational analyses (sampled to 1° × 1° resolution) from 1999 to 2004 are utilized here [although this model was renamed GFS from Aviation (AVN) in 2002, we identify it here as GFS regardless of the year]. The set of 12 hourly operational GFS model analyses collected in near–real time and used here is not complete. Thus, it is likely that additional ST occurred in this period.
b. Compositing methodology
NCEP–NCAR 2.5° × 2.5° monthly-mean reanalysis fields (Kalnay et al. 1996) for the period 1968–96 are used to define the mean climate structure in constructing the composite anomalies. Anomalies between the 1° × 1° GFS operational analyses and these long-term averages are calculated on the 2.5° × 2.5° grid. The anomaly composites reveal more detailed environmental features than are evident in the composited synoptic fields alone.
3. A framework for diagnosing ST occurrence
In compiling an ST climatology for the period 1957–2002 (Part II), it became evident that there was not a generally accepted conceptual model for ST. Candidate storms from previous studies (HURDAT information available online at http://www.nhc.noaa.gov/pastall.shtml#hurdat; Roth 2002) had a variety of structures and were not readily recognizable as a coherent class of storms. Hence, analyses of the 18 ST cases presented here were used to develop a novel set of criteria for classifying a storm as subtropical.
Subtropical storms are hybrid cyclones with cold upper-tropospheric and warm lower-tropospheric thermal anomalies (leading to the expectation of a relatively weak midlevel wind signature). ST are spawned as baroclinic developments in the presence of positive low-level vorticity and in an environment of relatively low static stability. Observations of wind structure support the diagnosis of ST as hybrid cyclones. For example, ascent data and model analyses from the developing phase of Karen (2001) reveal a near-surface cyclone of ∼30 m s−1 decreasing with height to approximately 3000 m (roughly 700 hPa) and cyclonic wind speeds increasing aloft with jets evident at 4000 and 7000 m (approximately 600 and 400 hPa, respectively; Fig. 1).
In a typical ST such as Karen, this hybrid structure remains evident until the convection erodes the upper-vorticity maximum to the point when the lower-tropospheric warm core dominates the system and the storm becomes more tropical in nature (e.g., Davis and Bosart 2003). Should this convective focusing of potential vorticity (PV) continue, this evolutionary path may lead the ST to undergo a tropical transition (Davis and Bosart 2004) to a tropical cyclone (TC) with a deep symmetric warm core, an upper-anticyclonic outflow, and convection encircling the surface low. Conversely, if the convection is not sustained, the cyclone may become more extratropical in character, typically leading to a less intense surface circulation than is the case in the tropical cyclone scenario (e.g., Hart 2003). Further support for the hybrid structure of ST comes from examination of both Advanced Microwave Sounding Unit (AMSU) analyses and model-derived structure diagnostics of tropical and subtropical cyclones.
One of the major justifications for the study of ST is that these storms are associated with damaging winds—and the public expect warnings on such systems. Indeed, the motivating case of Karen (2001) was a vivid demonstration to the second author of the impacts of this public perception! Although marine gale and storm warnings (for wind speeds greater than 17 and 25 m s−1, respectively) may be issued under World Meteorological Organization (WMO) guidelines (U.S. Navy 1994), the issuance of tropical cyclone warnings conveys a level of urgency to the public that is not typical of their response to other forms of storm warnings.
The rapid development and uncertainty associated with ST evolution provide a forecasting challenge for ST, which typically have large regions of gale-force winds (Guishard 2006). Given the societal importance of surface gales, a storm is classified as subtropical at the time when near-surface gales are first detected in the GFS analyses. In order for a cyclone to be classified as an ST, gales must be sustained for at least 36 h.
4. Summary and application of the ST criteria
A summary of the ST criteria developed based on the observations above and an elaboration on the methods of calculation of parameters used to characterize the ST and its environment are presented here.
a. Criteria for the identification of a subtropical cyclone
Given that a low pressure center has been tracked for at least 36 h in the model analyses or reanalyses, the criteria for classification as an ST are as follows:
the system attains gale-force winds (>17 m s−1) at 925 hPa1 and these gales must be sustained for at least three consecutive 12-hourly model analyses (equivalent to 36 h). The time of first onset of gales is defined as the ST formation time, T;
hybrid structure also persists for at least 36 h (i.e., more than one diurnal cycle). This hybrid structure criterion is determined using the CPS parameters (Hart 2003; also described later in this section);
only storms that form (i.e., attain gales) between 20° and 40°N are retained;
the cyclone should not have been tracked as either a purely cold- or warm-cored structure for more than 24 h prior to attaining hybrid structure; and
only storms located over the ocean from the first instance of a closed low through all instances of hybrid characteristics and the first occurrence of gales are considered.
The reasoning behind each of these criteria is as follows: the first criterion, the requirement for gales, is used to define storms likely to be sustained over a period of days and to have potential societal impact. Criteria 2 and 3 are applied to reduce the possibility of tropical and extratropical cyclones being included in the dataset. However, another process that results in a cyclone with hybrid structure in the latitude belt identified in criterion 3 is the extratropical transition (ET) of an initially tropical cyclone (e.g., Jones et al. 2003). ET typically results in a hybrid phase lasting 1–2 days (Evans and Hart 2003). Thus, criterion 4 was included to distinguish ST subtropical cyclones from ET events. Finally, criterion 5 was invoked because the presence of land may introduce factors such as differential surface friction and topography, which complicate the understanding of processes dictating subtropical cyclogenesis. A description of the methods used to calculate each of these ST criteria follows.
b. Selection and calculation of storm and environmental diagnostics
Previous investigations into the nature of tropical cyclogenesis (Ooyama 1963; Gray 1968; McBride and Zehr 1981; Emanuel 1986, 1995) and tropical transition (Bracken and Bosart 2000; Davis and Bosart 2003, 2004; Stewart and Bourassa 2007) have identified warm sea surface temperature (SST) or relatively high potential intensity as favorable for the development of a deep warm core associated with deep convection. Thus, SST is identified as a potential diagnostic of ST development. To assess the role of convection in ST development directly, two stability measures are also considered: the lifted index (LI) and convective available potential energy (CAPE).
Tropical cyclogenesis and development have been shown to be adversely affected by high shear (Gray 1968; McBride and Zehr 1981), whereas shear is needed for baroclinic cyclogenesis. Hart and Evans (2001) diagnosed regions where baroclinic cyclogenesis and tropical cyclogenesis were active to determine the likelihood of ET using shear [included in Hart and Evans (2001) via Eady baroclinic growth rate] and SST. Because ST are also hybrid structures, shear is included as a diagnostic parameter for ST formation.
The CPS (Hart 2003) parameters provide an assessment of the strength of the cyclone temperature anomaly based on height perturbation gradients (analogous to the thermal wind) and thermal asymmetry. Application of the CPS diagnostic to a large and diverse set of cyclones (real-time global CPS diagnostics are available online at http://moe.met.fsu.edu/cyclonephase/) demonstrates a continuum of cyclone types, typified by regimes of asymmetry or symmetry and core temperature anomaly (Hart 2003).
Examination of CPS diagnostics for ST Ana (2003) and Hurricane Fabian (2003) demonstrates the application of the CPS in distinguishing storm structure types. Atlantic ST have warm-core signatures extending only from the surface to the midtroposphere (Fig. 2a; Guishard 2006). Even during the most intense period of Ana as a 25 m s−1 tropical storm (Lawrence et al. 2005), the GFS-derived CPS only corresponds to a “moderate” lower-tropospheric warm core. In contrast, CPS analyses for Hurricane Fabian (Fig. 2b) signify a pronounced deep warm core from 31 August to 8 September 2003 typical of a mature tropical cyclone. For the majority of this period, Fabian was classified as a major hurricane of category 3 or 4 (Lawrence et al. 2005).
We make use of all of these parameters in the identification, characterization, and partitioning of ST. The individual parameters are calculated as follows:
Near-surface gales are a requirement for ST genesis in this analysis. The presence of gale-force winds and the radius to which they extend from the storm center are estimated for eight 45° storm-centered sectors in the GFS analyses, at 925 hPa. The algorithm used (Higgs 2005) searches through each 45° sector in radial increments of 50 km and includes those segments containing gale-force winds (>17 m s−1) in the calculation. The sector gale radii are then averaged and an area of gale-force winds is also calculated. A sector will be included in the calculation regardless of the extent of gales within that segment. This constitutes a consistent overestimation of the area covered by gales. However, in light of the fact that the model analysis resolution is 1° × 1°, this method of estimation is not unreasonable, because the real storm wind speeds are systematically underestimated by averaging over grid boxes.
SST diagnostics are calculated by taking a 2° × 2° average of the GFS SST centered on the cyclone.
- The CPS parameters are calculated over a 5° radius circle following Hart (2003): Evans and Hart 2003). Thresholds of −|VTL| > −10 and −|VTU| < 10 are used here to define hybrid structure.
c. Identification of the 18 ST cases presented here
A preliminary list of ST candidates was developed from an operational forecast-based catalog of subtropical storms for the period 1999–2003 (Roth 2002). The ST criteria detailed above were applied to identify ST among these candidate storms and to capture any additional storms in the dataset of 3455 available GFS hurricane season analyses from 1999 to 2004. Of the 30 storms in the Roth (2002) catalog, 15 ST events were identified. Three additional storms were found in the available GFS analyses during 2003 and 2004: Subtropical Storm Nicole (2004), Tropical Storm Otto (2004), and Tropical Storm Peter (2003). Characteristics of these 18 ST cases are reviewed in section 5.
5. Warm-season Atlantic ST, 1999–2004
The GFS analyses available for this study were from previous studies of tropical cyclogenesis and extratropical transition (Hart and Evans 2001; Hart 2003; Arnott et al. 2004), during North Atlantic hurricane seasons. There is the potential that additional ST cases have not been included in this survey, because only named tropical cyclone events of interest to the previous studies were archived in real time. Because a climatology of the annual distribution of ST (Part II) demonstrates a strong preference for North Atlantic ST events during the hurricane season, this warm-season focus is reasonable.
Eighteen examples of cyclones that meet the ST criteria (section 4) were identified in this GFS analysis set. The average genesis location of the 18 ST is 29.0°N, 62.2°W with 17 of these cases forming in a box bounded by 25°–35°N and 30°–80°W (Table 1; Fig. 3). The majority (14/18) of these ST genesis events is clustered between early August and mid-November, when SSTs are still relatively warm. Encroachment of upper troughs this far south is favored during the middle-to-late hurricane season (Postel and Hitchman 1999), rendering this region of the North Atlantic particularly favorable for subtropical cyclogenesis (Guishard 2006).
We now review the storm and environment characteristics for the 18 ST in this set to identify common elements in their genesis and subsequent evolution.
a. Gale radius
The societal impact of gale-force winds associated with ST cases such as the incipient Hurricane Karen (2001) was the initial motivation for this study. The range of gale-force wind radii varies widely across this ST database, from as little as 55 km up to over 1500 km. However the majority of ST cases have gale radii of a few hundred kilometers, with 50% between 180 and 540 km. In a typical ST, the gale radius expands with time (e.g., Hurricane Michael in 2000, as depicted in Fig. 4).
For 13/18 cases, ST genesis occurred over SST > 26°C; indeed, the average SST for the 18 cases is 27.2°C. Of the five storms (Ana, Noel, Olga, Otto, and Peter) that first attained gales over cooler waters (23°–26°C), three evolved in shear regimes hostile to tropical cyclone development (Table 1). No ST formed over SST warmer than the 29.7°C of Hurricane Gabrielle (2001).
c. Structure representation in the CPS
The mean CPS parameters at ST genesis indicate warm lower-tropospheric (900–600 hPa) and cold upper-tropospheric (600–300 hPa) thermal anomalies, corresponding to a hybrid structure with −VTL > 0 and −VTU < 0 (Table 1). The lower-tropospheric thermal asymmetry parameter is relatively symmetric (average B ≈ 8 m) compared to other cyclone types (Evans and Hart 2003), suggesting that the symmetric structure of these systems does not vastly depart from that of tropical cyclones. This region in the CPS (Fig. 5a) corresponds to a lower-tropospheric CPS regime typical of weak tropical cyclones. The thermal anomaly regime (depicted using −VTL and −VTL) of these 18 ST is that of a lower warm-core–upper cold-core structure (Fig. 5b). Only Florence, Kyle, and Otto (referred to as 3, 8, and 15 in Table 1 and subsequent figures) have substantially differing structure at genesis: Florence and Kyle had deep warm cores and Otto had a deep cold core. However, these three storms all had hybrid structures for at least 36 h immediately prior to attaining these structures, which confirms their status as ST.
d. Vertical wind shear
The average 900–200-hPa shear magnitude at the first occurrence of gales is 13.9 m s−1 with 10/18 evolving in shear exceeding 10 m s−1 over this layer. Hurricane Michael (2000) had the largest environmental shear (28.8 m s−1) at the onset of gale-force winds, consistent with its fast forward motion (Franklin et al. 2001). Many authors (e.g., Frank and Ritchie 2001; Elsberry and Jeffries 1996) have demonstrated that strong vertical wind shear (>10 m s−1 over 850–200 hPa) acts to weaken a developing tropical system through suppression of its core convection. Clearly, this restriction does not uniformly apply to ST (Table 1).
e. Vertical stability
The analysis of LI and CAPE reveals that ST development generally occurs in a region of reduced static stability. It is apparent that this drop in stability is due to the presence of a cold upper-cyclonic feature in close proximity to (if not superposed over) the surface cyclone. Comparison of an ST and a tropical storm for which we have direct observations illustrates the reduced static stability in the region of the ST genesis compared to the tropical storm environment.
Radiosonde soundings are available for both Hurricane Karen, which traversed Bermuda as an unnamed ST in October 2001, and Tropical Storm Bertha, a well-established warm-core system that tracked across Bermuda in July 2008 (Table 2). Similar temperature regimes are observed at the surface and 850 hPa for both storms, consistent with the CPS analyses of a warm core near the surface in both cases (as is typical of TC and ST). However, moisture and temperature profiles taken during the passage of Bertha directly over Bermuda are consistent with moist neutral stability (e.g., Emanuel 1986) with CAPE of 30 J kg−1 and LI of −0.09°C. In contrast, from sounding and stability indices measured during the passage of the precursor ST to Hurricane Karen over Bermuda, it is clear that the sounding taken is not that of a TC but something less stable: the CAPE in Karen is 770 J kg−1 with a lifted index of −1.6°C (Table 2), typical of a convectively unstable environment. CAPE and LI derived from GFS analyses for Karen (2001) are comparable to those derived directly from the sounding data, providing some measure of confidence in the model diagnostics of the convective potential of the systems.
In all 18 ST cases, the LI was negative and the CAPE exceeded 1000 J kg−1 at the time of the first presence of a closed low and/or onset of gales, indicating decreased stability in the ST genesis region. This further supports the notion of a cold upper-cyclonic feature and warm surface temperature destabilizing the column, giving rise to convection. Note that in addition, our analysis (not shown) reveals that the first occurrence of gales in each of the 18 ST cases is coincident or immediately following a sharp drop in CAPE and an increase in LI, which is consistent with the increase in winds as a direct response to the increase in convection.
6. Storm-centered composites
Storm-centered composites and composite anomalies of synoptic fields at different model pressure levels are derived from the GFS analyses for this 18-ST case study database. To ensure that the synoptic-scale features associated with ST genesis are captured in the analyses, the composites are created over a cyclone-centered 30° × 30° grid. This grid is monitored such that it does not extend across the equator: grid points in such regions are set to missing values. All storms are far from the poles, so no restriction is required on the northern edge of the composite domain. Composites have been produced for 12-h intervals from 48 h prior to genesis (T − 48 h) to 48 h afterward (T + 48 h). The anomalies represent departures from long-term NCEP–NCAR reanalysis (Kalnay et al. 1996) monthly-mean values for the period 1968–96.
The presence of an upper trough in the westerlies is the dominant feature during the development of an ST (Figs. 6, 7): a 300-hPa geopotential height trough between 954 and 960 dam is evident in the composites (Fig. 7). In each ST case, a trough was evident to the west of the surface low and no more than 5° away, although its intensity varied. For example, the upper trough is deeper than 948 dam for Karen, Michael, Noel, and Olga. The orientation of the trough varies from strongly negative (e.g., Juan in 2003) to neutral (e.g., Ana in 2003) to strongly positive (e.g., Florence in 2000). Typically, the PV associated with this trough contracts in scale from 10° (on the 300-hPa surface) to less than 5° as it approaches the ST genesis location.
The composites and composite anomalies suggest that the formation of an ST requires an upper trough to move into the region of high SST for a baroclinic low to be initiated at the surface (Fig. 7). When tempered with the fact that most ST occur in September, October, and November and with the need for deep convection to develop the lower-tropospheric warm core, the role of warm SSTs in ST genesis is understandable.
Two separate PV maxima are evident in vertical cross sections of PV and potential temperature (Fig. 6): one aloft and one near the surface. The upper cold-PV maximum is associated with a lowering of the dynamic tropopause (the 2-PVU contour; 1PVU ≡ 1 × 10−6 K m2 kg−1 s−1), with isentropic ascent to its east (e.g., Hoskins et al. 1985). The PV and potential temperature cross sections also reveal positive PV and θ maxima associated with the surface cyclone, consistent with the concept of a warm-core lower-tropospheric cyclone and a cold upper cyclone.
Height anomaly composites (Figs. 7d–f) reveal a −2-dam anomaly with zonal extent of approximately 12° longitude at 300 hPa and centered 7° to the west of the central sea level pressure anomaly 24 h prior to the onset of gale-force winds (Fig. 7d). This upper cold feature in the composite anomaly fields represents an upper trough or cutoff low in the upper troposphere. In addition, there is a strong ridge (positive SLP and 300-hPa anomaly fields; Figs. 7d–f) to the north of the surface anomaly. This is likely a manifestation of the anticyclonic anomaly in the poleward portion of a Rex block, as described in McTaggart-Cowan et al. (2006). In addition, there is anticyclonic shear to the north of the surface disturbances. Naturally, there is more curvature (i.e., less zonal flow) in the region of the upper cyclone, so there is a relative maximum in the westerlies to the north, causing anticyclonic shear just north of the surface storm. This configuration is consistent with the life cycle 1 (LC1) pattern of Thorncroft et al. (1993).
It is worth noting that the cyclonic anomalies (aloft and near surface) both grow in amplitude and extent, as would be expected in a baroclinically unstable disturbance. At 12 h after the onset of gales (T + 12, not shown), the cyclonic anomaly in the 300-hPa field has reached its maximum deviation of −6 dam from the mean. Also at this time, the anomalous upper ridge has built to its maximum strength of +16 dam from the mean. At T + 24 (Figs. 7c,f), the surface storm anomaly does not fill and even strengthens somewhat.
7. Potential genesis mechanisms
Hurricane Karen (2001) caused much consternation when gale warnings associated with the subtropical evolution of the storm near Bermuda went largely unheeded by the Bermudian public (Guishard et al. 2007). Its development in a baroclinic environment had caused NHC to delay in naming it (Beven et al. 2003), yet this ST genesis ultimately resulted in the development of a category 1 hurricane. Cases such as Karen have motivated research into tropical cyclogenesis via initially baroclinic processes (i.e., subtropical cyclogenesis).
Davis and Bosart (2003) present a theory for the development of North Atlantic tropical cyclones in which the mechanism for initiating a low–mid-level incipient vortex is baroclinic cyclogenesis in a vertically sheared environment. The resulting cyclone facilitates convection by inducing large-scale ascent and organizes the convection by low-level moisture convergence; convection weakens the vertical wind shear and a low-level vortex is spun up via the diabatic import of PV. It is the importance of this convective feedback on the low-level vortex that separates these storms from purely extratropical cyclones.
For the process described in Davis and Bosart (2003) to occur, a cold upper-cyclonic feature must penetrate equatorward into the subtropics from the midlatitude westerlies. Rossby wave breaking along trough axes embedded in a larger-scale anticyclonic shear is one mechanism for producing cutoff lows from an unstable, elongated, and positively tilted trough (LC1; Thorncroft et al. 1993). The synoptic-scale upper trough cuts off and reduces to a meso–α scale (Thunis and Bornstein 1996). The scale of the cutoff system developed via this mechanism is consistent with that typically observed for subtropical cyclogenesis.
In their study of a warm occluded cyclone over the northeastern United States, Posselt and Martin (2004) conclude that there is a significant contribution to Rossby wave breaking by the release of latent heat associated with convection on the frontal boundary of an occlusion. The effect is to facilitate Rossby wave breaking through development of a “notch” of minimum PV in the Rossby wave trough by diabatic dilution of PV. Assuming maintenance of a near-surface diabatic forcing (warm SST and a moist boundary layer), this convectively driven Rossby wave breaking could lead to ST development and ultimately to tropical cyclogenesis (Davis and Bosart 2003). These results support the findings of an earlier climatology (Postel and Hitchman 1999) that documented a maximum frequency of occurrence of Rossby wave breaking over the summer subtropical oceans.
The early evolution of Hurricane Karen (2001) demonstrates such a path to ST development. In this case, an equatorward PV fold (Figs. 8a,b) and the resulting reversal of the meridional PV gradient are favorable for the development of the cutoff cyclone that ultimately became Karen (Figs. 8c,d). The role of convection in the evolution of Karen is supported by soundings from Bermuda indicating CAPE of 770 J kg−1 and negative LI (Fig. 1a, 0000 UTC 12 October 2001, the same time as Fig. 8c). The symmetric structure and organized convection (Fig. 8d) preceded the naming of Karen 18 h later (Beven et al. 2003).
8. Environmental classification of subtropical cyclogenesis
One of the goals of this research is to provide a context for ST development of use to the forecasting community. Thus, the different ST environments are partitioned into four classes (Fig. 9; Table 1) based upon their shear2 and SST characteristics: tropical (SST ≥ 25°C; shear ≤ 10 m s−1); subtropical (SST ≥ 25°C; shear > 10 m s−1); the classical extratropical type 1 (E1; SST < 25°C; shear > 10 m s−1); and the low-shear extratropical type 2 (E2; SST < 25°C; shear ≤ 10 m s−1). The Hebert and Poteat (1975) satellite classification scheme was used to assign subtropical storm intensities via their ST number scale. The satellite signature at the time of genesis for each ST in this category is reproduced on the flowchart in Fig. 9. The relative paucity of extratropical genesis environments is due to the focus on the hurricane season in this study (cf. with the climatology of Part II).
a. Tropical environment (SST ≥ 25°C; shear ≤ 10 m s−1)
Seven of the 18 cases (39%) develop gales in a tropical environment (Table 1; Fig. 9; cf. 24% in Part II). All of the storms in this category formed in the later months of the Atlantic hurricane season. This warm-SST, low-shear environment is the most favored of the four partitions for tropical cyclogenesis. These SST and shear thresholds are symptomatic of an actively convecting region (Ooyama 1963; Gray 1968; Davis and Bosart 2003). Not surprisingly then, this partition results in the largest number of subtropical-to-tropical transitions (all seven of these cases).
Synoptic analyses of each of these case studies revealed the approach of an upper trough in the westerlies prior to ST genesis. Although forming in a relatively low-shear environment, all of these storms exhibited a further reduction in shear prior to the onset of gale-force winds (not shown). Analyses of the convective potential (LI and CAPE) of the background environment reveal that the CAPE (LI) typically increased (decreased) prior to the onset of gales as expected with an increase in active convection. This active convection and reduction in shear are consistent with the subtropical development described by Davis and Bosart (2003).
b. Subtropical environment (SST ≥ 25°C; shear > 10 m s−1)
The subtropical environment is characterized by a near-surface thermodynamic forcing due to heat and moisture fluxes from the warm ocean below in an environment of moderate shear due to an approaching baroclinic system. Seven of the eighteen ST (39%; cf. 33% in Part II) formed in this subtropical environment, with five of these ultimately becoming named tropical storms (Table 1).
A baroclinic system in the upper westerlies approaching the eventual genesis location will force mass ascent from the surface near the point of inflection between the upshear trough and the downshear ridge (Davis and Bosart 2003). This places the maximum vertical wind shear above the center of the surface anomaly, a configuration conducive to baroclinic development. In the presence of this forced ascent, the warm moist near-surface layer is susceptible to the generation of deep convection, bringing the column into near neutrality. In this near-neutral environment, the dynamic depth of the trough is increased and forced ascent can extend down to the surface. Convective feedback in this situation can lead to ST genesis and potentially also to tropical cyclogenesis.
c. E1 (SST < 25°C; shear > 10 m s−1)
Both the tropical and subtropical environments described above are characterized by warm SST. Because SST outside of the tropics is typically below the 25°C threshold chosen, we refer to this SST regime as the extratropical environment. We further partition the extratropical environment into high-shear (>10 m s−1 in the 900–200-hPa layer; E1) and low-shear (E2) regimes.
In a region of relatively cool SST, the approach of an upper trough yields an environment that is generally more (less) conducive to extratropical (tropical) developments. Even so, warm-air advection can enhance the potential for convection, enabling the maintenance of a lower-tropospheric warm-cored system on the order of days (thus leading to ST classification). Only 3 of the 18 warm-season ST cases are categorized in this high-shear and cool-SST E1 partition (Table 1; Fig. 9), a much smaller percentage than the cool season (Part II). Two of these E1 cases went on to become named tropical storms.
Subtropical Storm Ana (2003) is particularly notable among these three E1 cases because of its development of gales on 17 April, five months prior to the mean ST genesis date, above an SST of 23.7°C (3.5°C below the mean SST for the 18 ST cases). The unusual nature of Ana may have affected forecast accuracy for this storm. The existence of a hybrid low with gales (fitting our definition of an ST) was detected in the GFS as early as 17 April, although NHC issued the first advisory on Subtropical Storm Ana on 21 April (NHC advisory archives; available online at http://www.nhc.noaa.gov/archive/2003/ANA.shtml). In the NHC postseason reanalysis, Ana was officially determined to have reached tropical storm status on 21 April (Lawrence et al. 2005). Admittedly, this would have been a difficult forecast to make, because classical tropical cyclogenesis occurs over much warmer water (e.g., Gray 1968).
d. E2 (SST < 25°C; shear ≤ 10 m s−1)
One system in this warm-season set (Hurricane Olga in 2001) formed in this E2 environment, with neither the warm SST symptomatic of tropical development nor the high shear conducive to baroclinic cyclogenesis.
Close examination of the initial evolution of Hurricane Olga (2001) reveals that the system developed in a higher-shear environment over SST of 24.4°C (typical of E1) but that the onset of gale-force winds at 925 hPa occurred after the shear had decreased to 8.13 m s−1 (below the E1 shear threshold). Using the example of Olga, it may appear that the E2 partition is not really distinct from E1; however significant differences are evident in the composite signature of the trough in the E1 and E2 environments (Guishard 2006). Furthermore, whereas E2 was also the smallest category in the climatology, 19% of ST formed in this environment over the 45 yr (Part II).
9. Discussion and conclusions
A database of 18 ST cases forming in the 1999–2004 hurricane seasons has been examined here. To be classified as an ST, these systems are required to retain hybrid structure (as characterized in the CPS) and gale-force winds for more than one diurnal cycle. Seventeen of these formed in a region bounded by 25°–35°N and 30°–80°W (Fig. 3). Anomaly composites confirm the intrusion of an upper trough in the westerlies into a region of relatively warm SST (roughly 23°–27°C) and weak static stability (characterized through CAPE and LI) as the dominant pathway to the development of an ST (Fig. 7). Typically, the PV associated with the upper trough contracts in scale to less than 10° latitude as it approaches the cyclogenesis location (Figs. 6a,b), with a corresponding reduction in the environmental shear near the time of ST genesis. These conditions correspond well with the conditions for tropical transition identified by Davis and Bosart (2004). Dynamically, this process is analogous to the mechanism for the survival and ultimate reintensification of a TC in the face of a potentially destructive upper-PV anomaly through the diabatic redistribution of PV via convection: as part of this process, vertical shear is reduced over the surface cyclone (Molinari et al. 1998).
Two necessary (but not sufficient) conditions for tropical cyclogenesis are warm SST and weak vertical wind shear over the incipient disturbance. Consistent with their propensity to continue development into a tropical cyclone, 14 of the 18 ST cases formed over SST in excess of 25°C and 15 of the 18 went on to become named tropical cyclones. However, the shear constraint is less apparent for ST formation: the mean 900–200-hPa environmental shear at the time of genesis is 13.9 m s−1 with 10 of the 18 cases forming in an environment with shear in excess of 10 m s−1. This relatively high-shear environment corresponds to equatorward intrusion of upper troughs over the relatively warm SST present in the mid–late hurricane season. Storms such as Hurricane Michael (2000) that form in this environment and move equatorward from their genesis location often undergo tropical transition (Davis and Bosart 2004; Table 1).
As suggested by their genesis environments and the propensity for warm-season ST to evolve into tropical cyclones, the ST cases are partitioned into four subsets based on their environmental shear and SST. Contrasting the partition of genesis environments of these case studies to a larger climatology (Part II) reveals a similar distribution for the hurricane season. The warm-SST environment partitions (T and ST) have 39% each (7/18) of the ST database (cf. 33% and 24%, respectively, in Part II). The remaining four storms are split between the two cool-SST partitions: three are classified as E1 (strong shear), whereas only Olga (2001) is assigned to the E2 (shear ≤ 10 m s−1) partition. The E1 and E2 partitions were also less populous in the Part II climatology: of the 197 total ST identified in Part II, 23% and 19% were assigned to E1 and E2, respectively, with a majority of these occurring in the cool season.
The major motivation of the work presented here is to provide insights into ST genesis and to facilitate forecasting of such storms (Guishard 2006). Guishard et al. (2007) have applied these results to the region of Bermuda—the landfall location of the initial storms that motivated this research. The analyses presented here generalize the Guishard et al. (2007) study to the hurricane season in the wider North Atlantic region.
We appreciate discussion with Justin Arnott, Lixion Avila, Jack Beven, Lance Bosart, Bob Hart, Jessica Arnoldy (Higgs), Chris Landsea, Adam Moyer, and Aaron Pratt on this research. This work was supported by the National Science Foundation under Grants ATM-0351926 and ATM-0735973. The second author is grateful to the Bermuda Weather Service; BAS-Serco, Ltd.; and the Government of Bermuda Ministry of Transport and Tourism for their support during his doctoral thesis and to the Fessenden-Trott Trust for some supplementary support.
* Current affiliation: Bermuda Weather Service, St. George’s Island, Bermuda.
Corresponding author address: Jenni L. Evans, Department of Meteorology, The Pennsylvania State University, 503 Walker Building, University Park, PA 16802. Email: firstname.lastname@example.org
Recall that “shear” refers to the vector wind difference between 900 and 200 hPa.