Several recent landfalling tropical cyclones (e.g., Dennis, Floyd, and Irene 1999) have highlighted a need for a refinement in the forecasting paradigms and techniques in the area of quantitative precipitation forecasting. Floyd proved to be a particularly challenging forecast problem as it was accompanied by catastrophic flooding over large regions of the East Coast, in spite of its relatively quick northward movement. The extent and intensity of the precipitation distribution was strongly modulated by the storm's interaction with a midlatitude trough. In an attempt to better understand and quantify the relevant dynamics during this interaction, potential vorticity (PV) and quasigeostrophic perspectives are utilized.
As Floyd approached the East Coast, precipitation shifted to the left of the storm track due to the presence of a deep midlatitude trough in the Ohio valley. The juxtaposition of a cold-core PV anomaly associated with the midlatitude trough and a warm-core PV anomaly associated with Floyd created a strong and tropospheric-deep baroclinic zone along the eastern seaboard. This baroclinic zone provided a region favorable for deep isentropic ascent as the circulation of Floyd approached, resulting in prolific precipitation production. The latent heat release associated with this precipitation in turn acted to enhance outflow ridging north of Floyd, which was underpredicted by current numerical models. The enhanced outflow ridge resulted in enhanced jet-streak dynamics and a restructuring of the tilt of the midlatitude trough in a manner favorable for excessive precipitation production. Furthermore, the uplifting of the dynamic tropopause in southwesterly flow ahead of Floyd in response to ascent and differential diabatic heating resulted in a tropopause fold, a feature usually associated with upper-level fronts and differential subsidence in northwesterly flow.
Hurricane Floyd made landfall near Cape Fear, North Carolina, at 0900 UTC 16 September 1999 (09Z/16), as a category 2 hurricane on the Saffir–Simpson scale (Simpson and Riehl 1981) with maximum sustained winds near 50 m s−1. Most of the warnings associated with Floyd as it approached the coastline emphasized the potential for wind damage, since conventional wisdom indicated that the threat of catastrophic flooding would be minimized as Floyd accelerated to the north due to an interaction with a midlatitude trough. However, over 50 cm of rain fell in isolated areas of North Carolina, with widespread amounts exceeding 20 cm stretching from the Carolina Piedmont into southeastern New York (Fig. 1) (Lawrence et al. 2001). As a result of the extensive flooding, 56 people lost their lives in the United States and property damage was estimated at between $3 and $6 billion (more information available online at www.nhc.noaa.gov). Given the catastrophic nature of the flooding associated with Floyd, this paper will attempt to diagnose the dynamics associated with the extratropical transition (ET) of Floyd, which in this paper is loosely defined as transition from a relatively equivalent barotropic structure to a baroclinic or highly sheared environment. Special attention will be given to how these dynamics relate to precipitation distribution and intensity.
Past climatologies (e.g., Foley and Hanstrom 1994; Klein et al. 2000; Hart and Evans 2001) indicate that ET is a relatively common phenomenon in many subtropical ocean basins. Foley and Hanstrom (1994) performed an 82-yr (1909–90) climatology of tropical cyclones off of the west coast of Australia. Statistics from their study following the advent of satellite technology (1964–90) document 13 cases of ET or cyclone capture out of a total of 39 tropical cyclones (TCs). Composites of these 13 cases reveal that a key precursor to ET includes a highly meridional trough–ridge pair over the eastern South Indian Ocean. They also suggest that the juxtaposition of tropical and midlatitude air masses can often produce enhanced jet streaks, a feature discussed in many case studies of ET.
Klein et al. (2000) performed a 5-yr study (1994–98) of ET in the western North Pacific. They documented 30 cases of ET out of 112 tropical storms. A three-stage conceptual model of ET is proposed where 1) the interaction of the TC with a baroclinic zone leads to a temperature advection dipole, producing 2) strong ascent/descent over steeply sloped isentropes, resulting in 3) the decay and tilt of the warm core aloft and frontogenesis (Fig. 2, reproduced from Klein et al. 2000). Harr and Elsberry (2000) and Harr et al. (2000) related specific flow regimes and frontogenesis patterns to strong versus weak ET as it pertains to the above conceptual model. A region of broad slantwise ascent favoring the production of significant precipitation to the northeast of the TC is a signature of strong ET. This flow regime favors diabatic ridging ahead of the TC, increasing the amplitude of the trough–ridge couplet and leading to an enhanced jet streak poleward of the TC. These features are favored when the main midlatitude trough was located northwest of the tropical cyclone. Many of these features have also recently been documented by Sinclair (2002) in extratropical transition cases in the South Pacific.
A climatology of ET in the North Atlantic Basin performed by Hart and Evans (2001) indicated that ET was a more frequent occurrence (relative to the number of TCs) here than in the North Pacific. Out of 463 tropical storms documented from 1950 to 1996, 46% underwent some form of ET. The east coast of the United States and the Canadian Maritimes are the regions most likely to experience transitioning storms, with a frequency of 1–2 yr−1 over the 47-yr period. Fifty-one percent of all storms undergoing ET experienced posttransition intensification, most of which originated in the deep Tropics across the various ocean basins. A common ET signature includes a meridional trough–ridge couplet located to the west and poleward of the TC.
c. Case studies
Considering the frequency of ET, relatively few case studies have been performed. Sekioka (1956, 1957) first documented cases of ET in the Sea of Japan, noting that the presence of a pronounced baroclinic zone was necessary for ET to occur. Palmén (1958) and Anthes (1990) examined the large-scale dynamics associated with the ET of Hurricane Hazel (1954) along the east coast of the United States. Their studies concluded that the transport of tropical air masses into the midlatitudes resulted in a large increase in the available potential energy (APE) of the atmosphere. The approaching midlatitude trough provided a mechanism by which this APE was converted to kinetic energy through thermally direct circulations, resulting in a rather powerful extratropical cyclone.
Bosart and Carr (1978), DiMego and Bosart (1982a,b), and Bosart and Dean (1991) discussed the precipitation distribution and dynamics associated with the transition of Hurricane Agnes (1972). The first of these papers focused on flooding rains over Wellsville, New York, several hundred kilometers ahead of the tropical cyclone. The heavy rains were located in a confluent flow region between a weak midlatitude trough and the outflow ridge associated with Agnes. DiMego and Bosart (1982a,b) found that Agnes reintensified and then underwent ET as large-scale ascent ahead of the approaching trough overspread the periphery of the remnant low-level circulation. Bosart and Dean (1991) looked at the mesoscale aspects of the ET of Agnes and showed that precipitation was enhanced east of the Appalachians due to a modified cold-air damming event, with comparatively low pressure in the Southeast. Cool air trapped along the slopes of the mountains provided an environment favorable for isentropic ascent in easterly flow. This process is fairly common with landfalling tropical cyclones and was evident with Floyd (1999), as will be demonstrated later in this paper.
However, a transitioned cyclone does not have to be intense in order to have significant impact, as flooding can accompany relatively weak remnant cyclones. Hurricane Camille (1969) produced catastrophic flooding in the Virginia Piedmont as a relatively weak transitioned cyclone (Chien and Smith 1977). The heaviest precipitation fell to the north of a concentrated thermal ribbon in a region of warm-air advection. Bosart and Lackmann (1995) documented that David (1979) reintensified over land just prior to undergoing ET as it interacted with a relatively weak midlatitude trough. The ET occurred as the dynamic tropopause lifted ahead of the midlatitude trough due to diabatic heating from deep convection associated with David. The combination of enhanced ridging and downstream jet development “energized” the weak trough as positive potential vorticity advection (PPVA), roughly indicative of increasing cyclonic vorticity advection with height (CVA), increased over David and downstream in the equatorward-entrance region of the jet resulting in heavy precipitation downstream of the storm. Sinclair (1993) performed a case study of the ET of Tropical Cyclone Patsy (1986) finding that the greatest ascent in the redevelopment of Patsy also occurred in a region of an upward increase in CVA in the equatorward entrance region of a subtropical jet.
d. Recent research
The “PV thinking” perspective (Hoskins et al. 1985; Morgan and Nielsen-Gammon 1998) has proved to be useful in studies of hurricane–trough interactions (e.g., Bosart and Lackmann 1995; Molinari et al. 1995, 1997; Bosart et al. 2000) and transitioning tropical cyclones (e.g., Browning et al. 2000; Henderson et al. 1999; Thorncroft and Jones 2000; McTaggart-Cowan et al. 2001). The use of a PV perspective in these studies has revealed characteristic circulation signatures associated with lateral and vertical interactions among PV anomalies, and has established the importance of matching hurricane and trough scales to facilitate these interactions. This perspective has also indicated how transitioning tropical cyclones are able to maintain a warm-core structure as they progress over colder sea surface temperatures, and has quantified the importance of upper- and lower-level PV anomalies in the transition process.
A related issue is how to quantify the transition process. Hart (2003) and Hart and Evans (2003) have proposed a definition of transition based on a mapping of the horizontal thickness perturbations in the 850–600-hPa layer across the storm on one axis, and a measure of the height perturbation in the 900–700-hPa layer, relative to the storm environment, on the other axis. Their methodology yields a phase–space diagram, which can be used to map the life cycle of a storm in terms of its barotropic and baroclinic structure. More recently, Jones et al. (2003, manuscript submitted to Wea. Forecasting) performed a comprehensive review of the transition process utilizing a PV perspective. Here we will take a somewhat different, but related, approach in that we will interpret the Floyd (1999) transition from both a PV and a Sutcliffe (1947) and Sutcliffe and Forsdyke (1950) (advection of vorticity by the thermal wind) perspective. This approach will enable us to see that the ET process, in the case of Floyd, is gradual and not abrupt.
The main focus of this paper will be to understand the synoptic-scale dynamics associated with the transition of Floyd, with an emphasis on the precipitation distribution. What makes Floyd such an interesting case study is the challenge that it presents to conventional forecast wisdom relating precipitation amounts to cyclone speed. In order to understand the precipitation distribution, we will adopt a PV (see Hoskins et al. 1985; Morgan and Neilsen-Gammon 1998) and quasigeostrophic (QG) perspective. The paper is organized as follows. Section 2 contains the data and methodology, section 3 presents the results of the diagnostics, and section 4 summarizes and discusses the significance of these results.
2. Data and methodology
All gridded data used are taken from the National Centers for Environmental Prediction (NCEP) Eta Model twice-daily initialized grids at a resolution of 22 km. To ensure that the results presented are not strongly dependent on the resolution or initialization schemes of the Eta, a comparison of the PV fields taken from the NCEP Aviation Model (AVN) grids (at a resolution of 2.5° × 2.5°) versus those taken from the NCEP Eta model are shown in Fig. 3 for 12Z/16 and 12Z/17. Figures 3a and 3c are taken from the Eta grids while Figs. 3b and 3d are taken from the AVN model grids and show the evolution of the upper- and lower-level PV anomalies. Note that the maximum of low-level PV is indicative of a warm-core system (Floyd). While there are some minor differences between the two models in their representation of Floyd (e.g., the elongation of the lower-level PV maximum associated with Floyd in the Eta Model at 12Z/16), the general strength, scale, and location of the PV anomalies are quite similar. This provides some confidence that the model representation of Floyd is not scale dependent. It is important to note here that the general synoptic-scale environment of the storm is what is of interest for the purposes of this paper. While it is understood that representations of the core of Floyd in the model initializations are grossly inadequate, none of the assertions made in this paper are dependent on a concise representation of the storm core. Surface and upper-air station data are taken from the NCEP radiosonde and surface observing stations. All calculations and analyses are performed and displayed using the General Meteorological Package (GEMPAK) version 5.4 (Koch et al. 1983).
a. Synoptic overview
A comprehensive description of Floyd can be found in Lawrence et al. (2001). Briefly, Floyd achieved tropical storm status around 06Z/09 about 1200 km east of the Leeward Islands and became a hurricane 2 days later, approximately 300 km northeast of the Leeward Islands. Between 12 and 14 September, maximum sustained winds increased to 67 m s−1, placing Floyd at the top end of category 4 on the Saffir–Simpson scale (Table 1). The strength of Floyd and its proximity to the Florida coastline caused hurricane watches/warnings to be posted along a considerable stretch of the Florida east coast by late on 13 September. However, as a midlatitude trough approached the eastern seaboard, Floyd turned to the right, paralleling the Florida coastline. Early on 15 September, Floyd passed 150 km east of Cape Canaveral as it accelerated to the north (Fig. 1). Hurricane watches/warnings were then extended along a large stretch of the Southeast coastline from northern Florida to southern Virginia. By this time, heavy rain had already started to fall in the Piedmont region of the Carolinas.
As Floyd interacted with the rather intense midlatitude trough approaching the East Coast, heavy amounts of precipitation were deposited to the left of the storm track (Fig. 1). Over 40 cm of rain fell in southeastern North Carolina, with reports of over 30 cm of rain common throughout large portions of the east-central region of the state. In contrast to the very heavy amounts of precipitation to the left of the storm track, less than 6 cm of rain fell over the Tidewater region of Virginia to the east of the storm track.
A map of the 1000–500-hPa thickness and mean sea level pressure taken from the Eta initialized fields is given in Fig. 4. The minimum central pressure of Floyd is initially overestimated while still a tropical system on 00Z/16 (Fig. 4a) by some 40 hPa (950 versus 990 hPa). However, Floyd clearly has a warm-core structure at this time, as the low pressure center is collocated with a 582-dam thickness maximum. Meanwhile, a relatively intense high pressure system can be found centered near Sioux City, Iowa, with a central pressure of approximately 1021 hPa. The strong pressure gradient between the two systems is driving brisk northerlies and supporting cold-air advection into portions of the Midwest and Ohio valley. Meanwhile, a region of warm-air advection is already evident over the Carolinas, extending north up the coast into southeastern New York. This signature is consistent with the plume of clouds shown in the satellite picture in Fig. 1.
By 12Z/16, the gradient between the high pressure system and Floyd has increased in the model fields as the scale of Floyd begins to grow (Fig. 4b). By this time, the center of circulation is just beginning to undergo displacement to the left of the thickness ridge. Meanwhile, the thickness trough in the Midwest has shifted to the east, and its tilt has shifted from positive to neutral. Strong warm-air advection continues to spread north, as high pressure builds north of Floyd, driving strong geostrophic easterlies into New England. Cold-air advection is now evident much closer to the center of Floyd over the Appalachians and western Carolinas. As the system becomes more baroclinic, the Eta drops the central pressure of Floyd to 988 hPa, even as the best track data (Table 1) indicates that Floyd has filled to 967 hPa.
As Floyd continues north up the coast, the shape of the thickness field starts to take on a classic “S”-shaped configuration often associated with vigorous midlatitude cyclogenesis. The spatial scale of the storm expands as tightly packed closed isobars extend from Maine to North Carolina (Fig. 4c). An axis of cold-air advection stretches from the Ohio valley into the mid-Atlantic. The circulation center at this time still lies relatively close to the maximum thickness located near Long Island, consistent with its continuing classification as a tropical storm. However, by 12Z/17 (Fig. 4d) the circulation center of Floyd seems to be embedded within the thickness gradient. The magnitude of the thickness change across the circulation of Floyd is 24 dam, representative of a mean temperature change of 12°C.
b. A PV perspective
1) Horizontal maps
Figure 5 shows the upper- and lower-level PV and winds over the 36-h period ending 12Z/17. This representation is particularly useful in the diagnosis of ET, as the tropical system is marked by a maximum of low-level PV with low-PV air in the upper troposphere. Cold-core systems (midlatitude troughs) necessarily must have a maximum of PV in the upper troposphere. Because of this dichotomy in location of maximum PV, the interaction of tropical systems with midlatitude troughs can be simply displayed on a dual-level PV map.
At 00Z/16, Floyd is represented by a relatively symmetric maximum in the 850–700-hPa layer average PV (value of 1.4 PVU in Fig. 5a). Both the low- and upper-level winds are blowing cyclonically around the center of circulation, indicating that the core of Floyd remains isolated from the main baroclinic zone even as a large midlatitude trough approaches from the northwest. A relatively strong jet is located on the front side of this trough in the southwesterly flow with wind speeds approaching 75 m s−1. Over the next 12 h Floyd accelerates northward as suggested by a southerly wind of 12 m s−1 in the upper troposphere oriented along the PV gradient of Floyd at 12Z/16 (Fig. 5b). Also, consistent with the decrease in central pressure in the model field, the maximum PV value of Floyd has increased to 1.8 PVU. As the midlatitude trough continues to approach Floyd, the forward progress of the trough essentially comes to a halt north of Floyd, while the southern extent of the trough continues to move eastward. This differential motion causes the tilt of the trough to change from a positive to a neutral tilt.
By 00Z/17, the scale of Floyd has increased considerably, and there seems to be a warm frontal signature evident in the low-level PV field to the north of the circulation center (Fig. 5c). The magnitude of the low-level PV continues to increase to a value of 2.0 PVU. Floyd is now clearly embedded underneath a strong southwesterly jet, with wind speeds approaching 35 m s−1 over the warm front. Anticyclonic flow over eastern Canada increases in response to the erosion of PV from strong diabatic heating. The near superposition of the upper- and lower-level PV anomalies evident by 12Z/17 marks the inherent baroclinicity of the system (Fig. 5d). This baroclinicity is evidenced by the thickness gradient in Fig. 4d and implied in the 40 m s−1 jet now situated over the circulation center of Floyd (Fig. 5d).
Positive advection of PV can be found both to the east of the midlatitude trough and to the north of Floyd by 12Z/16 (Fig. 5b). By this time, most of the precipitation is located in the north and west quadrants of Floyd. As the scale and intensity of the low-level PV anomaly increases concomitantly with an increase in the steering currents, the magnitude of the differential advection of PV increases dramatically, implying a sharp increase in the forcing for ascent (descent) in the northwest (southeast) quadrants of the storm.
2) Cross sections
The interaction and eventual superposition of the PV anomalies can be readily seen in cross sections of PV and potential temperature (Θ) as displayed in Fig. 6. The cross-section orientations are given by the thick black lines in Fig. 5 and are generally taken along the mean temperature gradient. The juxtaposition of the warm core associated with Floyd and the cold core associated with the midlatitude trough results in a strong baroclinic zone.
In Fig. 6a, the relatively upright and narrow PV tube located on the right side of the figure represents Floyd. The downward bulge in the isentropes in the 850–150-hPa layer indicates the warm-core nature of the system. Note the relatively low-PV air located in the upper troposphere above Floyd. The midlatitude trough is located several hundred kilometers to the northwest of Floyd and has a somewhat broader PV anomaly. The cold dome in the midtroposphere is relatively pronounced with the 315-K isentrope located at about 400 hPa. This isentrope slopes downward to about 700 hPa in the core of Floyd. The slope in the isentropes represents a region capable of strong isentropic ascent (descent) given an easterly (westerly) component to the circulation associated with Floyd.
As the distance between the upper- and lower-level PV anomalies decreases, a region of enhanced baroclinicity is created by 12Z/16 (Fig. 6b). Furthermore, as the two PV anomalies begin to interact, they each become displaced from the maximum θ perturbations (deviations from the along cross-section mean). The midlatitude trough is now located downstream of the center of the cold-core θ anomaly while Floyd becomes displaced upstream of the warm-core θ anomaly. By 00Z/17, it is evident that the PV anomalies are undergoing superposition, even if the process is not yet complete (Fig. 6c). The remnant θ perturbations now have an upshear tilt, and the maximum PV lies in the θ gradient between the maximum perturbations. It is also apparent by this time that the horizontal scale of Floyd has increased, and is close to the scale of the midlatitude trough. The low-PV air and low static stability lying directly overhead of the low-level circulation suggest continuing convection near the core of the storm. As Floyd continues up the coast, superposition is completed by 12Z/17. At this point, a single coherent vortex tube with a strong upshear tilt is evident. The PV gradient at the front edge of the vortex tube becomes more diffuse in the upper troposphere as upright convection gives way to stratiform precipitation as will be demonstrated later in this paper.
In Fig. 7, an alternative to the above cross sections is presented where lines of θ have been replaced by isopleths of absolute momentum (M) [M = y + fυ, where y is the distance along the coordinate axis, f is the Coriolis force, and υ is the wind speed normal to the cross section; see Schultz and Schumacher (1999) for a more complete discussion]. For an equivalent barotropic system, the M surfaces should be vertical. Initially, Floyd displays an equivalent barotropic structure typical of tropical systems (Fig. 7a). However, by 12Z/16, M surfaces can be traced from the core of Floyd to near the core of the midlatitude trough. This is suggestive of the beginning of ET as the system begins to experience a slight upshear tilt (Fig. 7b). The tilt of the M surfaces becomes more pronounced over the next 12 h, and it becomes clear that ET is well under way (Fig. 7c). By 12Z/17, M surfaces can be readily traced from the center of the low-level PV anomaly to the center of the upper-level PV anomaly (Fig. 7d), indicating a single coherent system.
3) Dynamic tropopause maps
Potential temperature on the dynamic tropopause (DT, defined by the 1.5-PVU surface) is displayed in Fig. 8. The utility of this representation is that θ is conserved on the DT in the absence of diabatic and frictional effects (Morgan and Nielsen-Gammon 1998). Furthermore, gradients of θ on the DT are illustrative of regions of enhanced wind speed (i.e., jet streams). Also mapped are the 700-hPa vertical velocities as a proxy for regions of maximum latent heating. At 00Z/16 the value of θ in the core of the midlatitude anomaly is approximately 320 K while θ has a value of over 360 K near Floyd. The maximum ascent has already been displaced from the core of the storm, and is located near Wilmington, North Carolina, at 12Z/16 (Fig. 8b). The advective signatures suggestive of the eventual change in the tilt of the midlatitude trough are already present. Winds in the southern part of the trough are blowing strongly across the isentropes, whereas further to the northeast, the winds are much stronger but are blowing along the isentropes.
As the region of maximum ascent propagates up the coast, the 360-K isentrope expands to the north and west, even while the 330-K isentrope moves east over Michigan (Fig. 8b). This results in the tightening of the θ gradient on the DT. As Floyd moves over southeastern New York, the eastward progress of the midlatitude trough is halted from the New York border northward, while an eastward bulge in the isentropes becomes evident just south of this region (Fig. 8c). Finally the isentropes take on an S-shaped configuration by 12Z/17 (Fig. 8d) on the DT, about 12 h after this configuration had become evident in the 1000–500-hPa thickness field (Fig. 4c).
While the above representation is illustrative of a strengthening upper-level front, and an increasingly dynamically active trough, it fails to capture regions where the DT has actually folded. In these regions, the strength/gradients of the associated upper-level fronts can be underestimated in the figure above. Furthermore, since these folds occur lower in the troposphere, they can represent regions of decreased static stability, increasing the influence of the associated PV anomaly in producing a circulation on the lower boundary. Classical tropopause folds are most often associated with cold-air advection and subsidence maximizing along the warm boundary in northwesterly flow (e.g., Reed 1955; Bosart 2003). The resulting advections of PV and θ in this region of subsidence contribute to the frontogenesis and can produce a protrusion of PV into the midtroposphere.
A strong tropopause fold is associated with the ET of Floyd and is illustrated in Fig. 9, a north–south cross section of potential vorticity taken at 00Z/17 (solid line in Fig. 8c). However, unlike most classic tropopause folds associated with sinking air in northwest flow, the fold in this case is located in a region of southwesterly flow, ascent, and strong diabatic heating. This diabatic heating is evidenced by both the 700-hPa ascent region shown in Fig. 8c and the nonconservative nature of the PV field shown in Fig. 5. The fold is created in a region of differential ascent, as the DT is lifted to the northeast of Floyd, steepening the gradient of the DT with respect to height. The juxtaposition of the low DT associated with the midlatitude trough and the high DT associated with the storm produces a PV “wall.” The steepening of the DT by differential ascent is a mechanism discussed by Wandashin et al. (2000) and displayed schematically in their Fig. 4. However, a dry model is used in Wandashin et al. (2000) and most of the vertical motions are forced by temperature advections. In the case of Floyd, vertical motions are aided by latent heat release, which are maximized in the southwesterly flow ahead of the trough. Furthermore, the outflow generated in the upper tropopause acts to create a region of vertical shear, necessary for the production of the fold (see Fig. 4 of Wandashin et al. 2000). The most significant difference is that the mechanisms responsible for the fold produce a fold in the upper troposphere as opposed to the midtroposphere.
The relationship between θ on the DT and jet streaks is illustrated by examining 250-hPa maps from the same time period. At 00Z/16, a strong southwesterly jet, with maximum wind speeds of over 75 m s−1 stretches from the Ohio valley into southeastern Canada. Large values of divergence, on the order of 6 × 10−5 s−1, can be found in the equatorward jet entrance region just ahead of Floyd (Fig. 10a). By 12Z/16, the region of maximum 250-hPa divergence has propagated northward, once again just ahead of Floyd (Fig. 10b). The proximity and strength of the upper-level divergence with respect to Floyd suggests that Floyd has become coupled to the jet-entrance region, consistent with the findings of Sinclair (1993).
The jet entrance region becomes bifurcated as Floyd moves over the western tip of Long Island (Fig. 10c). By this time, the maximum wind speed as indicated by the wind barbs has increased to 85 m s−1. Possible mechanisms for the bifurcation of the jet-entrance region at this time include the transport of low-momentum air from the lower troposphere to jet level or preferential enhancement of the jet on the eastern side of Floyd in response to the outflow, in essence creating two separate jet-entrance regions. The divergence maximum becomes more diffuse by 12Z/17 (Fig. 10d), as Floyd starts to propagate to the cold side of the jet. The weakening of the divergence corresponds to a decrease in the vertical ascent (and the resultant precipitation) associated with the storm. The displacement of Floyd from the thickness maximum (Fig. 4d) and the corresponding decrease in the divergence suggests that the divergence was largely driven by ridging associated with the warm-air advection and precipitation associated with Floyd.
Broadly speaking, ET often represents the juxtaposition of two very different air masses resulting in a pronounced temperature gradient. Imposing the circulation from the tropical system produces a mechanism for strong isentropic ascent that can lead to copious amounts of precipitation. The diabatic heating resulting from the precipitation then serves as a mechanism for changing the scales and orientations of the upper and lower-level PV anomalies. The concentration of the latent heat release to the north of Floyd essentially erodes the northern extent of the midlatitude trough, resulting in an increasingly negatively tilted trough. A negative trough configuration favors strong cyclogenesis through a strong increase in differential CVA.
c. Precipitation distribution
In this section, we will attempt to analyze the synoptic-scale mechanisms controlling the precipitation distribution associated with Floyd. From a QG perspective, forcing for ascent can be attributed to differential vorticity advection and the horizontal Laplacian of temperature advection, in the absence of diabatic and frictional effects as demonstrated by Trenberth (1978). Alternatively, this can be expressed as the advection of geostrophic absolute vorticity by the thermal wind (Sutcliffe 1947; Sutcliffe and Forsdyke 1950). While diabatic effects are certainly at work in the case of Floyd, precise quantification of these effects is beyond the scope of this paper. It is important to note the QG perspective is being adopted as a diagnostic and not a predictive tool. Qualitatively, the signature of latent heat release is depicted in the PV and DT maps. The response of the atmosphere to diabatic forcing is already present in the rearrangement of the mass field. It is the signatures of this rearrangement of the mass field and its effect on the synoptic-scale or QG forcing that are at issue here.
Figure 11 is a simple representation of the Sutcliffe approximation and displays the 1000–200-hPa thickness and the 700–400-hPa layer-averaged geostrophic absolute vorticity. Initially at 00Z/16, the disparate nature of the two systems can be readily observed by noting that the 700–400-hPa vorticity maximum in Hurricane Floyd is located in the 1000–200-hPa thickness ridge (1234 dam), while the vorticity maximum associated with the mid latitude trough is located near the thickness trough (Fig. 11a). The equivalent barotropic structure of Floyd gives rise to little if any advection of vorticity by the thermal wind, indicating that the precipitation in the Carolinas (Fig. 12a) at this time is occurring in conjunction with convective and/or other mesoscale processes not specifically accounted for by QG dynamics. By 12Z/16, the QG forcing for ascent increases dramatically as the thickness gradient between Floyd and the midlatitude trough tightens, and the tropical vorticity maximum becomes somewhat displaced from the thickness maximum (Fig. 11b). Forcing for ascent is now implied in portions of the eastern mid-Atlantic, in conjunction with Floyd, as well as portions of central Pennsylvania in conjunction with the midlatitude trough. The broad nature of the precipitation field (Fig. 12b) lends credence to the broad nature of the forcing.
The transition process appears to proceed quickly from this point on, and by 00Z/17 the vorticity maximum in the midlatitude trough is no longer readily distinguishable from the vorticity maximum associated with Floyd (Fig. 11c). Furthermore, the maximum vorticity (24 × 10−5 s−1) is now embedded in the tightening thickness gradient. The result is that the QG forcing for ascent has increased dramatically in strength, while at the same time decreasing in aerial coverage. This is also illustrated by the decrease in coverage of the precipitation (Fig. 12c), concomitant with an increase in radar-derived base reflectivities over central and southeastern New York. By 12Z/17, any attempt to distinguish the vorticity associated with the midlatitude trough from the vorticity associated with Floyd is impossible, and ET can loosely be considered complete (Fig. 11d). The region under the greatest QG forcing is now located in parts of Maine and eastern Canada, as the system quickly races to the north (Fig. 12d).
In Fig. 13, the 850-hPa cyclonic vorticity, isotherms, and temperature advection are displayed to depict the interaction of Floyd with near-surface boundaries. Even as Floyd approaches the coast at 00Z/16, a region of warm-air advection (WAA) is evident over eastern North Carolina (Fig. 13a). This WAA is the result of the circulation of Floyd starting to overspread a preexisting temperature gradient. While a cold-air damming signature does seem apparent in the Southeast, a fairly consistent southeastward-directed temperature gradient is located along the entire length of the eastern seaboard, indicating a deeper, broader phenomenon than shallow cold-air damming. The region of WAA is collocated with heavy precipitation in the North Carolina Piedmont region well ahead of the actual cyclone center. Note that at this time, radar reflectivities still enclosed the circulation center of Floyd (Fig. 12a).
However, the precipitation structure quickly changes as Floyd moves onshore and cold-air advection begins on the southern flank of the storm by 12Z/16 (Fig. 13b). At this time, temperature advections are on the order of 24°C day−1 to the north and south of the center of circulation. An elongation of the vorticity in the region of WAA is indicative of a strengthening warm front. Farther to the south, a weaker reflection of a cold front is apparent along the Carolina coastline. The heaviest precipitation (Fig. 12b) continues to fall along and just west of the strongest thermal gradient (coastal front) and is consistent with the results of Bosart and Dean (1991). Meanwhile, precipitation in the southern and eastern quadrants of the storm quickly dissipates in response to the cold air being wrapped into Floyd on its southern flank. The classic S-shape signature in the thermal fields again becomes evident by 00Z/17 (Fig. 13c). The heaviest precipitation continues to lie along the western edge of the strongest WAA (Fig. 12c). As Floyd races to the north, the associated gradients of vorticity begin to weaken as the scale of the circulation (defined here as the area enclosed by the 4 × 10−5 s−1 contour) continues to increase (Fig. 13d).
At the surface, a representation of station θ and weather displays a classic coastal front signature (Fig. 14). At 00Z/16, winds back sharply from the east to the north along the North Carolina coastline corresponding with a rapid decrease in θ from the mid 20s°C into the upper 10s°C (Fig. 14a). This cold air is at least somewhat diabatically generated by the clouds and precipitation as the air at the surface east of the Appalachians is colder than air over the Appalachians in spite of the fact that 1000–200-hPa thickness field decreases to the west (Fig. 11a). As Floyd moves onshore, this boundary becomes more defined due to strong convergence, with temperatures across it dropping as much as 8°C (Fig. 14b). The location of the surface boundary exhibits a one-to-one correspondence with the region of the strongest radar reflectivities. By 00Z/17, northwesterly winds are located over northwest New Jersey while easterly winds can be found along the Jersey shoreline (Fig. 14c). The convergence across the surface boundary seems to be acting as a mesoscale focusing mechanism for the heaviest rainbands in the presence of strong synoptic-scale forcing for ascent.
4. Concluding discussion
Forecasts of precipitation distribution associated with landfalling tropical systems are still problematic in spite of recent advances in forecast modeling and techniques. Several recent storms (Floyd, Dennis, and Irene 1999) have illustrated some of the shortcomings of the conventional wisdom of relating storm totals to rate of storm motion. The threat of flooding was emphasized in the case of Dennis due to its relatively slow rate of propagation, although the heaviest amounts fell in relatively isolated areas and were on the order of 10–15 cm. In contrast, the warnings issued as Floyd approached the coast emphasized the threat of wind damage as the forward speed of the storm was expected to increase. However, flooding associated with Floyd proved to be by far the greatest threat to both life and property as widespread amounts greater than 20 cm fell on parts of the mid-Atlantic with isolated amounts exceeding 40 cm falling in eastern North Carolina. Considering the impact that the ET process seemingly had on the precipitation distribution and intensity associated with Floyd, this case is examined in order to try and elucidate the relevant dynamics modulating the precipitation distribution.
As the upper-level PV anomaly in the midlatitude trough approaches the low-level PV anomaly associated with Floyd, the low-level PV anomaly seemingly responds by elongating in the direction of the prevailing wind shear, while increasing in its aerial extent. Molinari et al. (1995, 1997), Bosart and Lackmann (1995), and Bosart et al. (2000) discuss the dynamics associated with favorable hurricane–trough interactions. Essentially, a favorable hurricane–trough interaction is dependent on a scale matching between the hurricane and the trough. In this way, the detrimental effects of the shear associated with the trough eventually become minimized in relation to the positive effects of the influx of cyclonic momentum into the tropical system. In cases of ET, the scale of the midlatitude trough is generally considerably larger than the scale of the tropical system. Furthermore, ET usually occurs over land or relatively cool SSTs, limiting the ability of latent heat fluxes to drive convection near the core of the storm and limiting its ability to withstand the shear associated with the trough. Therefore, in cases of strong ET, the scale of the tropical storm often expands to match that of the midlatitude trough as the core of the storm tends to propagate toward higher values of shear. This evolution represents a change in the relevant dynamics of the storm from diabatic to baroclinic in nature. The effects that this evolution can have on the central pressure of the system are dependent on a number of factors and are beyond the scope of this paper.
In the case of Floyd, the dynamics evolved in a manner that accentuated the potential for heavy precipitation. The juxtaposition of the relatively cool dry air of the midlatitude trough and the warm moist air associated with Floyd produced an intense, troposphere-deep baroclinic zone. The circulation center of Floyd then interacted with this baroclinic zone, producing deep isentropic ascent and precipitation north of Floyd. The diabatic heating associated with the precipitation acted to erode the positive PV while the associated outflow acted to enhance the downstream ridging. The ridging then changed the orientation of the midlatitude trough from a positive to a negative tilt. The end result was a rather potent storm exhibiting both tropical and extratropical characteristics.
A by-product of the diabatically induced ridging is the creation of a tropopause fold. The DT is steepened in a region of strong differential ascent, which in this case is maximized in the southwesterly flow ahead of the trough. The differential ascent in combination with the juxtaposition of two very different air masses (one characterized by high-PV air in the upper troposphere, the other characterized by very low-PV air in the upper troposphere) acts to create a PV “wall.” This wall is then differentially advected into a tropopause fold as vertical shear is enhanced by the outflow associated with Floyd.
Another result of the juxtaposition of the midlatitude and tropical air masses is the enhancement of the jet streak on the east side of the midlatitude trough. This process is most aptly represented by the increase of θ on the DT. Alternatively, this process can be viewed as an enhancement of the outflow jet of Floyd as the trough approaches. The dramatic acceleration of the wind in the jet entrance region produces a region of strong divergence in the equatorward-entrance region. Floyd then becomes coupled to the jet-entrance region as it propagates up the coast.
The difficulty in producing an accurate numerical simulation of the transition process is perhaps related to the impact of the diabatic effects on the synoptic-scale mass field. Accurately portraying and predicting the structure of the outflow is seemingly crucial in correctly simulating the evolution of the transition process. However, the structure of the outflow associated with the tropical cyclone is heavily dependent on the strength and extent of the convection in the tropical storm, which is in turn dependent on model scale and parameterization schemes. Failures with regard to the precipitation distribution can be traced back to the models' inability to accurately build the ridge associated with the outflow of Floyd (Fig. 15). Figures 15a and 15b show the 48-h forecasts of the AVN and Eta Models in the 250-hPa height field verifying at 0000 17 September, while Fig. 15c shows the AVN initialization for 0000 UTC 17 September. Differences between the forecasts and the verification center around the ridge in the northeastern United States are shown in Fig. 15d. By failing to accurately predict the shape and intensity of the ridging in the eastern Atlantic, the models failed to simulate the sharpness of the trough–ridge couplet and the associated synoptic-scale forcing. Note that the AVN Model was chosen here because of its coverage into the Atlantic.
While accurate forecasts of ET may prove daunting, it is perhaps comforting to note that the dynamics of ET can be simply represented to the first order in a QG framework. Maps modeled after the Sutcliffe approximation to the QG omega equation provide a concise physical explanation of the evolution of the precipitation distribution during ET. Initially, most of the QG forcing for ascent is relegated to the front side of the midlatitude trough in a region of CVA by the thermal wind. However, as the circulation of Floyd becomes displaced from the thermal ridge, the region of maximum CVA by the thermal wind quickly transfers to the north and west quadrants of Floyd. Eventually, as the trough tilt becomes negative, the CVA by the thermal wind maximizes as the wavelength between the trough and downstream ridge shortens. The strong forcing for ascent in combination with the copious amounts of available moisture helps explain the excessive precipitation amounts accompanying the relatively quick propagation of Floyd along the East Coast.
From the QG perspective, the transition of Floyd can perhaps be quantified by the location of the vorticity maximum associated with Floyd relative to the tropospheric thickness gradient. Floyd starts being displaced from the warm core in the thickness field by 12Z/16, and is both thoroughly embedded in the thickness gradient as well as no longer distinguishable from the midlatitude trough by 12Z/17. From the PV perspective, the start of transition may be considered to begin when the lower-level PV anomaly starts to change scale at 12Z/16, and ends with the superposition of the upper- and lower-level PV anomalies by 12Z/17. It should be noted, however, that no dynamic quantification of transition is clean or precise given the continuous nature of the spectrum of storms in reality. The beginning and end of ET is difficult to precisely quantify, as the process is gradual and not binary in nature. As such, it is perhaps better to discuss ET in terms of a spectrum of storm structure.
While the results presented here concern only one storm, it should be noted that several cases of transition have been examined, and the results associated with Floyd are broadly applicable to cases of ET where the upstream trough is the dominant feature. In these cases, the precipitation shifts to the left of the storm track as illustrated by Irene in 1999. Subsequent papers will present both composite and specific case studies of several storms representing each class of landfalling tropical systems discussed above.
The authors wish to thank the NSF for Grants ATM-9912075 and ATM-0000673, which made this work possible. We are also deeply grateful for the help and support (technical and otherwise) of Dr. Frank Marks (Hurricane Research Division), Anantha Aiyyer, Dr. Michael Dickinson, Michael Cempa, Kristen Corbosiero, and Kelly Lombardo.
Corresponding author address: Eyad H. Atallah, Dept. of Earth and Atmospheric Sciences, University at Albany, State University of New York, 1400 Washington Ave., Albany, NY 12222. Email: firstname.lastname@example.org