The Garden City, Kansas, Storm during VORTEX 95. Part I: Overview of the Storm s Life Cycle and Mesocyclogenesis

Transcription

1 372 MONTHLY WEATHER REVIEW The Garden City, Kansas, Storm during VORTEX 95. Part I: Overview of the Storm s Life Cycle and Mesocyclogenesis ROGER M. WAKIMOTO, CHINGHWANG LIU, AND HUAQING CAI Department of Atmospheric Sciences, University of California, Los Angeles, Los Angeles, California (Manuscript received 23 July 1996, in final form 30 April 1997) ABSTRACT Analysis of a supercell storm that produced an F1 tornado near Garden City, Kansas, is presented. This event provided one of the first opportunities to synthesize data collected by a new airborne radar platform called ELDORA (Electra Doppler radar) developed by the National Center for Atmospheric Research. The early stages of development of the midlevel mesocyclone and the entire evolution of the low-level mesocyclone are captured over a 70-min period. The low-level mesocyclone began as an incipient shallow circulation along a synopticscale trough. The circulation intensified and grew in depth via vortex stretching under the influence of a strong updraft. As this rotation built up from the boundary layer, it initially remained separate and distinct from the midlevel mesocyclone. Subsequently, the two mesocyclones merge to produce a single column of rotation 4 5 km in diameter. An occlusion downdraft develops within the mesocyclone circulation during the last passes by the storm signaling the beginning of the tornadic phase. Perturbation pressure retrievals provide conclusive evidence that this downdraft is driven by a downward-directed pressure gradient force. 1. Introduction During the past few decades, conventional and Doppler radar studies have provided the only comprehensive documentation of the structure of tornadic storms. The former identified systematic radar reflectivity features (e.g., Stout and Huff 1953; Fujita 1958, 1981; Browning and Donaldson 1963; Browning 1964; Forbes 1981) and the latter revealed important kinematic patterns based on single- (e.g., Brown et al. 1978; Burgess and Lemon 1990; Burgess et al. 1993) and multi-doppler velocity analyses (e.g., Ray 1976; Brandes 1978, 1981; Heymsfield 1978; Ray et al. 1981; Johnson et al. 1987; Brandes et al. 1988) that have now become well known in the field. The availability of these datasets was timely as they were soon compared with early numerical simulations of supercells (e.g., Klemp et al. 1981). This coupling of observations with numerical simulations significantly advanced our understanding of the dynamics of severe storms. It is now known how the midlevel mesocyclone within the supercell is initiated and maintained. Updrafts within supercell thunderstorms usually begin to rotate as a result of tilting of low-level horizontal vorticity associated with strong vertical shear of the environmental winds (Klemp 1987). Subsequently, a separate Corresponding author address: Dr. Roger M. Wakimoto, Department of Atmospheric Sciences, UCLA, 405 Hilgard Ave., Los Angeles, CA roger@atmos.ucla.edu baroclinically generated low-level mesocyclone develops along the rear-flank gust front, stretches in the vertical owing to forced uplift along the front, and rapidly exceeds the vorticity of the midlevel rotation (Klemp and Rotunno 1983; Rotunno and Klemp 1985; Klemp 1987; Davies-Jones and Brooks 1993). The different mechanisms that generate these two circulations may explain the observational fact that, based on 20 years of Doppler radar operation in Oklahoma, only 30% 50% of mesocyclones are accompanied by tornadoes (Burgess et al. 1993). More recent data suggests that the percentage may be as low as 20% (D. Burgess 1997, personal communication). Brooks et al. (1993, 1994a,b) have shown that the presence of an intense rotation at midlevels of the storm does not guarantee that a persistent low-level mesocyclone will develop. Brooks et al. (1993) use their results to argue that the question of whether an environment will support tornadic thunderstorms is not the same as the question of whether it will support supercells. Indeed, there still does not exist an accepted theory on how supercell tornadoes form (Rasmussen et al. 1994) although there has been progress made on nonsupercell tornadogenesis (e.g., Brady and Szoke 1989; Wakimoto and Wilson 1989; Roberts and Wilson 1995; Lee and Wihelmson 1997). Contributing to this lack of understanding is the relatively small number of case studies on supercell storms. The entire archive of complete dual-doppler radar observations of supercells probably consists of fewer than 10 cases (Brandes 1993). This is not surprising given 1998 American Meteorological Society

2 FEBRUARY 1998 WAKIMOTO ET AL. 373 the difficulty in capturing these events within a relatively small, ground-based, dual-doppler network. A shortcoming of these pioneering radar studies was the relatively coarse temporal and spatial resolutions of the kinematic wind fields. Typical radar volume scans were separated by 9 10 min and horizontal grid resolutions were approximately 1 km (e.g., Brandes et al. 1988). Such resolutions were appropriate for determining the genesis and maintenance of the mesocyclone; however, understanding the processes that lead to tornadogenesis requires a higher resolution dataset. Recent numerical studies of supercells have already approached this finegrid resolution (120 m) that appears capable of simulating both the low- and midlevel mesocyclone and the tornado circulation (Wicker and Wilhelmson 1995). The movement of a supercell into and out of the dual- Doppler lobe defined by the radars [see Doviak et al. (1976) for radar scanning geometry] has made it difficult to collect a single dataset that captures both the incipient stages of midlevel rotation and the development of the tornado. Moreover, the horizon for a ground-based radar is such that even at 0 elevation angle the height of the beam above the ground will increase at distant ranges. This can result in low-elevation scans frequently collecting data above cloud base at distances greater than 80 km (e.g., Burgess et al. 1993; Doviak and Zrnić 1993). Accordingly, the kinematic storm structure near the ground, where spatial gradients are often the strongest, suffer from a lack of data from radars (Klemp 1987). This is the first of a two-part presentation that documents a tornadic event near Garden City, Kansas, on 16 May 1995 during the Verification of the Origins of Rotation Experiment (VORTEX). The focus of this study is the airborne Doppler radar data collected by the NCAR (National Center for Atmospheric Research) ELDORA (Electra Doppler radar) as it made numerous passes by the parent storm. High spatial resolution data was collected by ELDORA on the Garden City storm for over a 70-min period. This time span encompasses the early development of the midlevel mesocyclone and the entire evolution of the low-level mesocyclone. The transition from this larger scale rotation into an F1 tornado is also documented. The elevated radar platform and the close range of the flight tracks to the storm ( 15 km) results in an accurate synthesis of the nearground wind field. Part I presents an overview of the life cycle of the Garden City storm and focuses on the evolution of the low- and midlevel mesocyclones. Part II (Wakimoto and Liu 1998) examines the wall cloud pendant from cloud base and the development of the tornado. Section 2 briefly describes the VORTEX field experiment and the ELDORA platform. The environmental conditions and visible satellite imagery of the storm are presented in section 3. The radar methodology and an overview of the Garden City storm s life cycle are presented in sections 4 and 5. Section 6 focuses on the evolution of the low-level mesocyclone and section 7 examines the relationship between the low- and midlevel mesocyclone. The forcing mechanism of the occlusion downdraft is shown in section 8 and a summary and discussion is presented in section VORTEX field experiment and ELDORA a. VORTEX During the spring of 1994 and 1995, a large multiagency field project focused on studying issues concerning tornadoes and tornadic storms was operated within northern Texas, Oklahoma, and southern Kansas. The experimental design was unique with the primary data platform being approximately a dozen instrumented vehicles and the NOAA (National Oceanic and Atmospheric Administration) P-3 and NCAR Electra aircraft equipped with tail Doppler radars. The mobile nature of the experiment was intended to increase the target of opportunities and to allow for perfect positioning of vehicles and aircraft for collecting high quality data. Some of the vehicles were capable of making upper-air observations using the CLASS (Cross-chain Loran Atmospheric Sounding System). All vehicles were equipped with an instrument package recording vehicle position, velocity, and standard meteorological variables (temperature, humidity, pressure, wind direction, and speed) with a temporal resolution of 6 s (Straka et al. 1996). Other observational platforms included routine soundings from the National Weather Service (NWS), profiler data from the Forecast Systems Laboratory, surface observations from the NWS Automatic Surface Observing System (ASOS), the WSR-88D Doppler radar network (Klazura and Imy 1993), mobile 3-mm (Bluestein et al. 1995) and 3-cm (Wurman et al. 1996) wavelength Doppler radars, a portable 3-cm wavelength Doppler radar (Bluestein and Unruh 1989), and the Oklahoma mesonet (Brock et al. 1995). Detailed forecasts were provided each morning during the experiment in order to identify the geographic area with the highest potential for supercell development. Time of departure and target area for the instrumented vehicles and the aircraft were based on these forecasts. Several updated forecasts were provided later in the day and transmitted to VORTEX personnel while en route to refine the location for potential storm development. For more information on the experimental design of VORTEX, the reader is referred to Rasmussen et al. (1994). b. ELDORA Since the 1970s airborne radars have played an important role in advancing our understanding of storms that were either too remote or occurred too infrequently for ground-based radars to sample adequately (e.g., Jorgensen 1984; Marks et al. 1992; Wakimoto et al. 1995; Wakimoto and Atkins 1996). These airborne radars are able to provide high spatial resolution data by following

3 374 MONTHLY WEATHER REVIEW TABLE 1. ELDORA scanning mode. Antenna rotation rate ( s 1 ) Number of samples PRF (Hz) Number of range gates Gate length (m) Sweep angle resolution ( ) Along-track resolution (m) Maximum range (km) Individual unambiguous velocities ( m s 1 ) Maximum unambiguous velocity ( m s 1 ) / / a meteorological feature over an extended period of time. Accordingly, these platforms have provided an opportunity to collect a wealth of information on important phenomena such as hurricanes (e.g., Marks and Houze 1984), microbursts (Hildebrand and Mueller 1985), convective storms (Ray et al. 1985; Wakimoto and Atkins 1996), and fronts (Wakimoto et al. 1992). Recognizing the potential for this type of mobile platform, the National Center for Atmospheric Research (NCAR) and the Centre de Recherche en Physique de l Environnment Terrestre et Planetaire (CRPE) in Paris, France, developed a unique airborne Doppler radar with the following capabilities: 1) Increased accuracy and sensitivity is achieved by averaging more independent samples in the radar pulse volume thus decreasing the uncertainty of the Doppler velocity and reflectivity estimate (Doviak and Zrnić 1993). This significantly enhances the ability of the radar to detect motions in the clear air. 2) Higher spatial resolution in the along-track direction by using a faster rotation rate of the antenna. 3) Larger unambiguous dynamic range for velocity measurements to minimize the ambiguities often associated with severe convection (Doviak and Zrnić 1993). The result of this joint endeavor is ELDORA, which was flown for the first time during TOGA COARE (Tropical Ocean and Global Atmosphere Coupled Ocean Atmosphere Response Experiment; Webster and Lukas 1992) in While the results from the first field deployment of ELDORA have been promising (Hildebrand et al. 1994, 1996), the opportunity to test the system at its full capabilities did not occur until the spring of 1995 during VORTEX. Representative examples of its capabilities to collect data on convective and clear-air phenomena have been presented by Wakimoto et al. (1996). The choice of scanning parameters used during data collection on the 16 May storm is shown in Table 1. The interested reader is referred to Hildebrand et al. (1994, 1996) for discussion on the radar design of ELDORA. 3. Environmental conditions and satellite analysis The Garden City tornado formed at approximately 2330 UTC (hereafter all times in UTC, UTC CDT FIG. 1. (a) Upper-air sounding launched at 2311 UTC 16 May 1995 from Dodge City, Kansas. The thick black line represents the approximate path a surface-based parcel would take. Vertical profile of the wind is shown with the half-barb, full barb, and flag representing 2.5, 5.0, and 25.0 m s 1, respectively. (b) Storm-relative hodograph based on the wind profile at Dodge City. Storm motion was 15.4 m s 1 from 254. The location of this launch site relative to the Garden City storm is shown in Fig h) 16 May The sounding launched at Dodge City (located 75 km to the east-southeast) at 2311 revealed the strong convective instability within the area (Fig. 1a). The convective available potential energy (CAPE) was estimated to be 2903 J kg 1 with an equilibrium level at about 13 km AGL. The ELDORA radar measured storm tops up to 16 km AGL. Maximum wind speeds on the sounding were 65 m s 1 near 190 mb. The storm-relative hodograph (Garden City mesocyclone was translating at 15.4 m s 1 from 254 ) is shown in Fig. 1b. The favorable speed shear is apparent and can be categorized as a quasi-straight or unidirectional hodograph with a southwest to northeast orientation. Two other soundings were examined to assess the representativeness of the sounding shown in Fig. 1. The thermodynamic profile and wind hodograph based on a mobile sounding launched in Dodge City at 2339 was

4 FEBRUARY 1998 WAKIMOTO ET AL. 375 analogous to those shown in Fig. 1. Another sounding launched at Medicine Lodge at 2313 (about 220 km to the east-southeast of the storm) showed stronger clockwise turning of the winds at low levels and larger helicity; however, it was not chosen for two reasons: 1) the moisture and temperature profiles were different than those shown in Fig. 1 and the NCAR sounding at 2339, and 2) its geographic location was deemed too far to represent the environmental conditions. Considering the changes in a wind profile that can occur in a short time interval (e.g., Davies-Jones et al. 1990), however, it is possible that the hodograph shown in Fig. 1 may not properly characterize the environment. Results by Klemp and Wilhelmson (1978) and Rotunno and Klemp (1982) show how a thermal disturbance develops in an environment characterized by a wind shear vector that does not vary in direction with height. Initially the thermal will remain symmetric about the wind shear vector as it grows with two midlevel mesocyclones developing within the updraft. These mesocyclones are located to the right and left of the mean wind and rotate cyclonically and anticyclonically, respectively. Subsequently, these two mesocyclones split along with the updraft producing two mirror image storms. The storm-relative helicity is small based on the hodograph shown in Fig. 1b. Helicity has been tested as a forecast tool for tornadic activity with large values increasing the potential for strong to violent tornadoes (Davies-Jones et al. 1990; Moller et al. 1994). However, low helicity values by themselves cannot be used to rule out the possibility of tornadoes because helicity can be an extremely volatile parameter (Davies-Jones et al. 1990). The upper-air charts revealed a cutoff low at the 500- mb level located west of Baja California at May 1995 (not shown). The center of the low moved into central Arizona by May as it approached the VORTEX area. There was warm, moist air being advected into Kansas by strong winds at the 850-mb level in advance of the trough. A key feature at the 250- mb level was the location of the jet axis, which appeared to be centered over Dodge City at May. The Kansas Colorado state border was located in the forward left quadrant of the jet streak. These upper-level conditions combined with the sounding launched at Dodge City suggest that severe convection was possible over western Kansas and eastern Colorado (e.g., Miller 1967; Schaefer 1986). The surface analysis at 2100 is shown in Fig. 2. Out ahead of a southward propagating cold front from Canada is a trough passing through Iowa, Kansas, southeast Colorado, and New Mexico. The trough is characterized by a wind shift from northeasterly to southwesterly flow through Kansas and a weak thermodynamic discontinuity. The dryline (separating relatively warm, dry air to the west from the cool, moist air to the east) extends FIG. 2. Surface analysis at 2100 UTC 16 May Temperature, dewpoint temperature, and wind direction and speed are plotted (wind speed notation same as in Fig. 1). Isobars and a trough are indicated by solid and dashed black lines, respectively. The position of a cold front and the dryline (drawn as a scalloped line) are shown. The black arrow denotes the position of the surface observations from Garden City, Kansas. northward and appears to intersect the trough near the Kansas Colorado border. Surface analyses superimposed on high-resolution visible satellite imagery at 2200, 2230, 2300, and 2330 UTC are shown in Fig. 3. Towering cumuli began developing at the intersection between the dryline and trough line (not shown) around 2130 near the Kansas Colorado border. Deeper convection became apparent at this location by 2200 (Fig. 3). Note the cloud line indicating the position of the trough in southeast Colorado at this time. The Electra began data collection on the southern part of the storm at The convective activity intensified as the storm began to propagate northeastward along the trough at 2230 and The winds at Garden City are southeasterly at 2300 suggesting inflow into the storm. These winds switch to southwesterly and increase in magnitude in response to the outflow winds at 2330 in Fig. 3. The tornado formed about 10 km east of Garden City a few minutes after this last image. 4. ELDORA flight track and airborne radar methodology The movement of the Garden City storm based on data collected by the ELDORA tail radar data is shown in Fig. 4. The first indication of a possible hook echo begins at with a band of echo extending to the southeast away from the main storm. The appendage is well developed by as it moves south of Garden City. The black line located north of the airport is the tornado damage track (to be shown in more detail in Part II). Note that a weak-echo eye (Fujita 1958, 1965, 1981; Zrnić et al. 1985; Wakimoto and Atkins 1996) in

5 376 MONTHLY WEATHER REVIEW FIG. 3. Surface data superimposed on high-resolution visible satellite imagery at 2200, 2230, 2300, 2330 UTC 16 May Temperature, dewpoint temperature, and wind direction and speed (wind speed notation same as in Fig. 1) are plotted for selected sites. The position of a trough and dryline are shown by the black dashed and scalloped lines, respectively. State boundaries are drawn as black lines. The positions of the surface observations at Garden City (GCK) and Dodge City (DDC) are labeled on the images. Dodge City also denotes the location of the WSR-88D site. FIG. 4. Evolution of the Garden City echo at 600 m AGL based on radar reflectivity data collected by ELDORA. The time interval (UTC) for each echo composite is labeled on the figure. The flight track of ELDORA is also shown. Radar reflectivity is drawn as gray lines with values greater than 40 and 50 dbz hatched and shaded gray, respectively. The tornado track is indicated by the black line located north of the Garden City airport. The site where a picture of the tornado (shown in Fig. 2 in Part II) was taken is shown by the black cross located southeast of the airport. The site where a picture of the wall cloud and blowing dust (shown in Fig. 8 in Part II) is shown by the black cross located south of Garden City.

6 FEBRUARY 1998 WAKIMOTO ET AL. 377 TABLE 2. ELDORA analysis times. 2220: :40 UTC 2226: :05 UTC 2234: :45 UTC 2242: :30 UTC 2247: :30 UTC 2301: :00 UTC 2308: :42 UTC 2313: :20 UTC 2319: :00 UTC 2324: :30 UTC 2331: :00 UTC the middle of the hook echo is seen centered over the beginning of this track at This relationship is described further in part II. Also shown in Fig. 4 is the approximate race track pattern flown by the Electra to the south of the storm. The aircraft passes were nominally 5 6 min long and were primarily oriented northeast southwest. A major objective of VORTEX was to collect high resolution data of the mesocyclone and the tornado. This requirement led to flight plans that were near low altitudes ( 300 m AGL) and relatively close ranges to the hook echo/mesocyclone (10 15 km). Data collection commenced nearly 70 min before tornadogenesis. Table 2 lists the times of the flight legs that were analyzed in this paper. Dual-Doppler syntheses were not completed for the times after the tornado developed (i.e., after 2336). The vortex after this time was too narrow (see part II) for the radar beam to accurately resolve the circulation. A description of the methodology used to create the kinematic wind fields from ELDORA and an error analysis are presented in the appendix. At the distances from the echo depicted in Fig. 4, it was nearly impossible to synthesize the kinematic wind field in the upper parts of the storm owing to the high elevation angles of the radar. Accordingly, the dual- Doppler analyses shown in this paper were restricted to a height of 5 km AGL and the vertical velocities were obtained from the anelastic continuity equation by upward integration of the horizontal convergence field. This height was adequate to assess the evolution of the midlevel mesocyclone. 5. Overview of the storm s life cycle Horizontal cross sections at 0.6 and 3.4 km AGL (hereafter, all heights are AGL) of radar reflectivity, vertical velocity, and vertical vorticity are shown in Fig. 5. The first pass by the developing storm occurred at (Fig. 5a). The position of a fine line, bounded by the 0-dBZ isopleths, delineates the subsynoptic trough 1 discussed in Figs. 2 and 3. Fine lines are areas of enhanced echoes that are often seen with a sensitive Doppler radar at locations of surface convergence boundaries (e.g., Wilson and Schreiber 1986; Wilson et al. 1994). Although the storm initiated along this boundary, the main precipitation echo is located approximately 10 km to the northwest. The storm-relative velocity field at 0.6 km indicates uniform flow from the northeast with no prominent surface feature. There are downdrafts less than 5ms 1 near the 50-dBZ echo on the southwest part of the echo and a component of the main updraft into the storm is located several kilometers to the east. A small pocket of vorticity in excess of s 1 is apparent along the trough. The wind field and the vorticity analysis at midlevels reveal the cyclonic and anticyclonic mesocyclones centered on maxima in updrafts (Klemp 1987). The orientation of these mesocyclones relative to the mean hodograph shown in Fig. 1b is consistent with past studies (Klemp and Wilhelmson 1978). The main updraft at low levels has intensified to greater than 7.5 m s 1 by (Fig. 5b). The area and intensity of the downdrafts within the main echo has increased and results in a northerly component in the flow near the main precipitation echo compared to Fig. 5a. The magnitude of the vorticity along the trough has also increased and an arrow denotes the location of a shallow vortex that is the incipient low-level mesocyclone. Small pockets of updrafts are seen along the trough axis. The counterrotating mesocyclones are still evident at midlevels in both the wind and vorticity fields. The major changes are that the updraft speeds and vorticity within the cyclonic mesocyclone have increased to greater than 30 m s 1 and s 1, respectively. There is an enhanced area of northerly flow at low levels owing to the downdrafts as seen in Fig. 5c. In addition, there has been a slight veering in the direction of the environmental wind field south of the fine line (trough). This suggests enhancement of the inflow into the growing storm. The increased northerly flow combined with the inflow results in an expansion in the area influenced by updraft, which now has a component over the low-level vorticity maximum. The initial low-level rotation has deepened under the influence of this updraft and now extends up to 3.4 km. The vorticity analysis at this level reveals two distinct maxima associated with the cyclonic midlevel mesocyclone and what was originally the incipient low-level mesocyclone. The persis- 1 There is some uncertainty whether this fine line delineates the trough since there were a number of lines detected by the WSR-88D on this day. The in situ thermodynamic measurements recorded by the aircraft (not shown) during penetrations of the line, however, are consistent with the surface observations shown in Fig. 3. Moreover, an examination of the time sequence of high-resolution satellite images suggest that the Garden City storm initiated along a cloud line that was associated with the trough.

7 378 MONTHLY WEATHER REVIEW FIG. 5. Horizontal cross sections at 0.6 and 3.4 km AGL of radar reflectivity, storm-relative velocity vectors, vertical velocity, and vertical vorticity at (a) 2220: :40, (b) 2234: :45, (c) 2242: :30, (d) 2247: :30, (e) 2308: :42, and (f) 2324: :30 UTC. Radar reflectivity are shown as gray lines with values greater than 40 dbz shaded gray. Positive and negative vertical velocities are shown as solid and dashed gray lines, respectively. Positive and negative vertical vorticity values are shown as solid and dashed black lines, respectively. For (a) (e), the contour interval for vertical velocity at 0.6 and 3.4 km is 2.5 and 10 m s 1, respectively. The contour interval for the vertical vorticity for (a) (e) is s 1. Owing to the intense gradients that develop in (f), the contour interval for vertical velocity at 0.6 and 3.4 km is 5 and 20 m s 1, respectively. The contour interval for the vertical vorticity for (f) is s 1. The dashed boxed-in areas shown in (e) and (f) are enlarged in Fig. 8. The location of vertical cross sections shown in Fig. 10 are shown by the thick black lines on the vertical velocity and vorticity plots at 0.6 km. The track and direction of the aircraft is shown on the radar reflectivity plots at 0.6 km.

8 FEBRUARY 1998 WAKIMOTO ET AL. 379 FIG. 5.(Continued) tence of the midlevel anticyclonic mesocyclone is not surprising given the quasi-linear hodograph shown in Fig. 1. Unlike past studies, however, the Garden City echo does not split producing mirror image storms (e.g., Charba and Sasaki 1971; Klemp and Wilhelmson 1978; Rotunno and Klemp 1982). The absence or relatively weak intensity of the left-moving supercell may be attributed to the storm moving north of the boundary where low-level inflow air is cooler. 2 2 Atkins and Weisman (1997, personal communication) have performed preliminary numerical simulations of the Garden City storm

9 380 MONTHLY WEATHER REVIEW The low-level mesocyclone intensifies at 0.6 km at (Fig. 5d) with the maximum situated along the gradient in the updraft. An appendage forms in response to precipitation particles being advected by the rear-flank downdraft. The trough has undergone a major transition by this time. The inflow into the storm and the developing downdrafts have resulted in the evolution of this boundary into the classical rear- and forwardflank gust fronts (e.g., Lemon and Doswell 1979). The s 1 isopleth of vorticity at low levels approximately encircles the location of these two gust fronts with the vorticity maxima situated at the intersection as suggested by past schematic models. The cyclonic mesocyclones have merged at midlevels. The anticyclonic mesocyclone is still evident in the vorticity analysis and the kinematic wind field although a small area of downdrafts is located near s 1 isopleth. The vorticity exceeds s 1 at low- and midlevels at in Fig. 5e. The updraft pattern has assumed an S shape at 0.6 km as forced uplift occurs along both gust fronts. The extension of the updraft to the southwest is well correlated with the fine line and convergent wind fields. The echo appendage is better defined at low levels and has assumed a hook shape at 3.4 km. Apparent at both levels is the beginning stages of a weakening of the updraft near the center of the cyclonic mesocyclone. This trend continues in Fig. 5f with a downdraft minimum less than 10 m s 1 surrounded by updraft at 0.6 km. The center of the mesocyclone still lies along the updraft downdraft gradient. The Garden City tornado develops by the next pass of the aircraft approximately 6 min later. 6. The low-level mesocyclone a. Origins of rotation The results in the preceding section (cf. Fig. 5a with 5f) suggest that the ELDORA dataset captures the better part of the life cycle of the low-level mesocyclone. In an attempt to determine the origins of this circulation, an examination of the four earliest passes by the Garden City storm was undertaken (Fig. 6). The analysis area in the figure was chosen to be approximately centered on the low-level mesocyclone. The fine line associated with the synoptic-scale trough is outlined by the 0-dBZ contours in Fig. 6a. An isochrone analysis of this feature revealed that its movement was 330 at 2.4 m s 1.To better document mesoscale features along the trough line, the trough-relative wind fields are shown in Fig. 6 rather than the storm-relative wind fields. It should using the hodograph shown in Fig. 1b. These experiments include the effects of a realistic frontal boundary instead of a homogeneous environment. Their results confirm that a weak left-moving storm develops as it propagates over the cold air north of boundary. be noted that a sizable fraction of the area shown in the figure requires syntheses of winds within the clear air. The increased sensitivity of ELDORA allowing for the detection of motions in the convective boundary layer, which is devoid of precipitation scatterers, was discussed by Wakimoto et al. (1996). The relative flow in Fig. 6a is consistent with the location of the trough with southerly flow in the southern section of the domain turning to easterly flow north of the fine line. Sections of the trough are characterized by vertical vorticity values greater than s 1. The updraft into the storm and the downdraft from precipitation is evident in the northwest section of the domain. A feature that was not apparent in Fig. 5 is a series of updraft maxima greater than 2 m s 1 along the trough. The forcing mechanism of this alongfrontal updraft structure is not known although there was a suggestion of horizontal convective rolls within the convective boundary layer seen in the radar reflectivity and Doppler velocity data (not shown). Their orientation appeared to be perpendicular to the trough line. Past studies suggest that horizontal convective rolls developing within the convective boundary layer can interact with frontal boundaries producing a periodic structure in updrafts (Wilson et al. 1992, 1994; Atkins et al. 1995). The periodic updrafts persist and grow in horizontal extent by the next pass of the aircraft shown in Fig. 6b. The vertical vorticity field has undergone a major change with the maxima now appearing within the vertical velocity gradients. The vorticity maxima enclosed by the s 1 isopleth along the trough was determined to be the incipient mesocyclone. As will be shown in section 7, this feature was relatively shallow ( 1 km) at this time. The relationship between the vertical velocity and vorticity along the trough line suggests that tilting of horizontal vorticity may play a role in their generation. Figure 7 represents an enlargement of the boxed-in area in Figs. 6a and 6b. The vertical velocity and vorticity analyses are superimposed on the horizontal vorticity vectors. The location of the trough is drawn on the figure and is coincident with a discontinuity of the vorticity vectors. South of the trough the vorticity vectors are disorganized in Fig. 7a but clearly indicate a southerly direction in Fig. 7b. This is expected based on the hodograph and the orientation of the cyclonic and anticyclonic mesocyclones at midlevels shown in Fig. 5 suggesting that these vectors represent the environmental vorticity. North of the trough the vectors are easterly, which is probably a result of the horizontal vorticity being produced by baroclinic effects. Recall that there was relatively cooler and drier air on the north side of the boundary (presented in Fig. 3 and verified by in situ measurements recorded by the aircraft). The gradients in vertical motion produce a pattern of negative and positive tilting of the baroclinically produced vorticity north of the trough axis in Fig. 7a. Also apparent is a small area of positive stretching. The mag-

10 FEBRUARY 1998 WAKIMOTO ET AL. 381 FIG. 6. Horizontal cross sections at 0.6 km AGL of radar reflectivity, trough-relative velocity vectors, vertical velocity, and vertical vorticity at (a) 2220: :40, (b) 2226: :05, (c) 2234: :45, and (d) 2242: :30 UTC. Radar reflectivity are shown as gray lines with values greater than 40 dbz shaded gray. Positive and negative vertical velocities are shown as solid and dashed gray lines, respectively. Positive and negative vertical vorticity values are shown as solid and dashed black lines, respectively. The dashed boxed-in areas shown in (a) and (b) are enlarged in Fig. 7. The track and direction of the aircraft is shown on the radar reflectivity plots at 0.6 km.

11 382 MONTHLY WEATHER REVIEW FIG. 7. Enlargement of the boxed-in areas shown in Figs. 6a and 6b. Vertical velocity, vertical vorticity, horizontal vorticity vectors (top figure), and the components of tilting and stretching from the vorticity equation (bottom figure) are shown at (a) 2220: :40 and (b) 2226: :05 UTC. Positive and negative vertical velocities are shown as solid and dashed gray lines, respectively. Positive and negative vertical vorticity values are shown as solid and dashed black lines, respectively. Positive and negative values of stretching are shown as solid and dashed gray lines, respectively. Positive and negative values of tilting are shown as solid and dashed black lines, respectively. The location of the synoptic-scale trough is shown by the long dashed black line. nitude of both the tilting and stretching increase during the next pass of the aircraft (Fig. 7b). The vorticity field has evolved into isolated pockets of cyclonic vorticity that are still situated between velocity maxima (also noted in Fig. 6b). The horizontal separation of these vorticity maxima is approximately equal to the spacing between areas of positive tilting in Figs. 7a and 7b providing evidence that this is the primary mechanism for the pattern. However, the approximate collocation of the stretching and vertical vorticity maxima in Fig. 7b (there is a small displacement between the maxima in stretching and the western vorticity maxima) suggests that the former is the dominant mechanism at this time. The absence of anticyclonic vorticity in Fig. 7b may result from the cyclonic wind shift along the trough axis (see Figs. 3 and 6). This pool of cyclonic vorticity would potentially mask a small anticyclonic vortex. The wind field in Fig. 6c indicates that cold air outflow (as evidenced by the northerly flow) from downdrafts generated along the rear flank has reached the western section of the trough. These outflow winds have enhanced the convergence along the trough axis resulting in stronger updrafts and vertical vorticity. The location of the incipient low-level mesocyclone is highlighted on the figure and it is associated with a cyclonic circulation in the wind field. The interval between Figs. 6c and 6d is important since it has been identified as the beginning stage when the synoptic-scale trough evolves into the rear- and forward-flank gust fronts. The wind field along with the vertical vorticity patterns have assumed a configuration similar to the schematic models shown by several investigators (e.g., Lemon and Doswell 1979). At the intersection of the rear- and forward-flank gust fronts is the location of the developing low-level mesocyclone with vorticity greater than s 1 and a circulation in the wind field. Results to be presented in section 7 show that the vorticity associated with the low-level mesocyclone has significantly deepened by this time. b. Mature stage leading up to tornadogenesis The rapid development of the low-level mesocyclone is illustrated by the vertical vorticity and velocity fields at 600 m shown in Fig. 8. The vorticity is greater than s 1 in Fig. 8a and is centered near the transition between updraft and downdraft (e.g., Rotunno 1981; Brandes 1993). The 5 m s 1 isopleth is elongated to the south and northeast in response to the enhanced forced lifting along the rear- and forward-flank gust fronts, respectively. Davies-Jones and Brooks (1993) have shown that baroclinic generation and tilting of vorticity in the rear-flank downdraft can be large at this time. The 10 m s 1 isopleth located north of the mesocyclone is a component of the main updraft into the storm. The gust fronts rotate cyclonically around the mesocyclone as the circulation intensifies in Fig. 8b. A major transition has occurred in the vertical velocities located on the eastern side of the mesocyclone. A small

12 FEBRUARY 1998 WAKIMOTO ET AL. 383 FIG. 8. Enlargement of the area centered around the low-level mesocyclone at (a) 2301: :00, (b) 2308: :42, (c) 2319: :00, and (d) 2324: :30 UTC. Storm-relative vectors are plotted. Vorticity values are drawn as black lines. Positive and negative vertical velocities are drawn as solid and dashed gray lines with the zero isopleth drawn as a thick gray line. The location of analysis shown in (b) and (d) are shown in Figs. 5e and 5f. Thick black lines indicate the location of vertical cross sections shown in Fig. 10. area of negative velocities is apparent (also noted in Fig. 5). This feature marks the beginning of the occlusion downdraft discussed in past numerical (Klemp and Rotunno 1983) and observational studies (e.g., Barnes 1978; Lemon and Doswell 1979; Brandes 1981, 1984a; Hane and Ray 1984; Carbone 1983). The occlusion downdraft has been hypothesized as a possible triggering mechanism for tornadogenesis although the details have not been clearly shown. The development of this downdraft starts min before tornadogenesis. The downdrafts intensify in Fig. 8c with velocities less than 5 ms 1. The forced uplift along the rearflank gust front and the occlusion downdraft produce an updraft structure that has assumed a horseshoeshaped pattern (e.g., Brandes 1978; Klemp and Rotunno 1983; Rotunno 1986; Wicker and Wilhelmson 1995). The negative vertical velocities extending to the south of the mesocyclone are related to the downward flow associated with the horizontal roll vortex within the head of the rear-flank gust front (e.g., Droegemeier and Wilhelmson 1987; Mahoney 1988). Past studies have shown that this stage is typically accompanied by an annular ring of vertical vorticity (e.g., Klemp and Rotunno 1983; Brandes 1993; Wicker and Wilhelmson 1995). There is a suggestion of an arc-shaped vorticity pattern in Fig. 8c; however, it is not prominent.

13 384 MONTHLY WEATHER REVIEW FIG. 9. Horizontal projection of the representative trajectories of air parcels superimposed onto the storm-relative wind field at 2308: :42 UTC and (a) vertical vorticity and (b) stretching of vorticity. Angles labeled on the figure represent the final location of the parcels along a circle with a radius of 1.5 km centered on the mesoscyclone. Time intervals of 5 min are labeled along the trajectories as black dots. The pass by the storm preceding tornadogenesis is shown in Fig. 8d. The rear-flank downdraft has intensified northwest of the mesocyclone also shown in the numerical simulations by Wicker and Wilhelmson (1995)] and is continuous with the occlusion downdraft located east of vorticity center. This downdraft pattern has been associated with the clear slot or clearing of the clouds spiraling to the south and east of the possible location of the tornado (e.g., Lemon and Doswell 1979). The position of the vorticity maximum located between the updraft and downdraft along the maximum gradient of vertical velocity has been shown before (e.g., Brandes 1993) although the spatial resolution shown in the figure is believed to be unprecedented. c. Trajectories The origins of the air parcels that comprise the lowlevel mesocyclone at were investigated. This was identified as the period when the low-level mesocyclone was undergoing rapid intensification. Parcels located every 5 along a circle with a radius of 1.5 km were centered on the vorticity maximum. Trajectories were determined by backward time integration using a fourth-order Runge Kutta method with a time step of 10 s. The dual-doppler velocity fields were linearly interpolated between analysis times before integration. Past attempts in the literature to reconstruct trajectories were forced to assume stationarity (e.g., Brandes 1981, 1984a, 1984b). The horizontal projection of eight of these trajectories are superimposed onto the vertical vorticity and the vertical component of vorticity stretching (Fig. 9). The angles labeled on the figure represent the final location of the parcels along a circle with a radius of 1.5 km centered on the mesocyclone. The stormrelative wind field at is superimposed onto the trajectories in Fig. 9. All of the parcel trajectories originate from the lowlevel (less than 1 km) environmental flow in the warm sector. Most of the parcels cross the forward-flank gust front into the cool air before spiraling into the mesocyclone from the north and northwest sectors (Fig. 9a). As the parcels near the forward-flank gust front they undergo intense vortex stretching as shown in Fig. 9b. Tilting was relatively small at this time with no clear trend toward positive values. It is suspected that positive tilting of horizontal vorticity did exist but was not resolvable in the radar analysis. These results confirm the hypotheses advanced by numerical simulations of supercells (e.g., Klemp and Rotunno 1983; Rotunno and Klemp 1985). Wicker and Wilhelmson (1995) identified another source region located at midlevels northwest of the mesocyclone. No parcels of air originated from this location in the present study; however, this may be attributed to the coarser vertical resolution in the present analysis compared to the higher spatial resolution in

14 FEBRUARY 1998 WAKIMOTO ET AL. 385 FIG. 10. Vertical cross sections of radar reflectivity, vertical velocity, and vertical vorticity through the low-level and midlevel mesocyclones at (a) 2234: :45, (b) 2242: :30, (c) 2247: :30, (d) 2308: :42, (e) 2319: :00, and (f) 2324: :30 UTC. Locations of vertical cross sections are shown in Fig. 5. Radar reflectivity are drawn as gray lines with values greater than 40 dbz shaded gray. Positive and negative values of vertical velocity and vorticity are drawn as gray lines and dashed gray lines, respectively. Vorticity and velocity values greater than s 1 and 30 m s 1 are shaded gray. Vertical velocities less than 30ms 1 are hatched. recent numerical simulations. In addition, the lowest data level in the present study was limited to 200 m. 7. Relationship between low-level and midlevel mesocyclones It was demonstrated in Fig. 5 that the midlevel mesocyclone was separated from the low-level mesocyclone by a horizontal distance of 8 10 km. This displacement is not inconsistent with the numerical simulations by Klemp and Rotunno (1983), Rotunno and Klemp (1985), Klemp (1987), and Davies-Jones and Brooks (1993). They show that two separate mechanisms are required to generate these circulations. Vertical cross sections illustrating the growth and merger of the mesocyclones in relation to the echo and vertical velocities are shown in Fig. 10. The vertical vorticity and velocity patterns aloft are correlated at (Fig. 10a) with the former exhibiting values greater than s 1 at midlevels. This relationship is expected since vortex stretching within the main updraft is the primary mechanism to intensify the midlevel mesocyclone. A developing weak-echo vault (e.g., Browning and Ludlam 1962) is seen in the reflectivity pattern. Also shown in the figure is the vorticity associated with the incipient low-level mesocyclone along the trough line. The s 1 isopleth suggests that this feature is approximately 1 km deep. The scenario shown in this figure is reminiscent of the separate upper-level and low-level isentropic potential vorticity anomalies discussed by Hoskins et al. (1985) for extratropical cyclogenesis. The area encompassed by the updraft expands by the next aircraft flyby (Fig. 10b), largely a result of forced uplift at the leading edge of the developing rear-flank downdraft (see Fig. 6d). The initial low-level vorticity

15 386 MONTHLY WEATHER REVIEW perturbation has rapidly deepened under the influence of these updrafts. It should be noted that the maximum vorticity may not be displayed at all levels in Fig. 10 owing to the difficulty of selecting a vertical cross section through coherent features that often tilt with height. The mesocyclones merge to produce one circulation at UTC (Fig. 10c) embedded within the main updraft of the storm. The classical echo overhang is seen in the reflectivity field. The scenario described in the figure is similar to a case described by Vasiloff (1993). He documents a tornado that formed along a low-level convergence boundary away from the mesocyclone core. With time, the mesocyclone intensified while the tornado seemed to migrate toward its center. Unfortunately, no analyses are presented so detailed comparisons are not possible. The vorticity increases to greater than s 1 in Fig. 10d ( ) and is no longer tilted with height. The updraft structure has undergone a major transition at this time. The positive vertical velocities exhibit a bimodal structure with a trough of weaker updrafts approximately at the location of the strong vorticity. This is the formative stage of the occlusion downdraft (Klemp and Rotunno 1983). The trend continues at (Fig. 10e) as a column of weak downdrafts between two updraft maxima lies within the mesocyclone, which has now intensified to greater than s 1. This cross section also reveals the horizontal vortex associated with the rear-flank gust front (see Fig. 5e). The backside of the rotating head within the gust front produces a separate area of downdrafts that were discussed in Figs. 8c and 8d. The mesocyclone rotation strengthens in Fig. 10f ( ) with the occlusion downdraft associated with speeds greater than 30 ms 1 at 4 km. Tornadogenesis occurs immediately after the analysis shown in this figure. FIG. 11. Retrieved perturbation pressure at 2324: :30 UTC superimposed onto the wind field at 600 m. Isobars are drawn as gray lines with values less than 2 mb shaded gray. The black line denotes the location of a vertical cross section shown in Fig Forcing mechanism of the occlusion downdraft The development of the axial downdraft within the mesocyclone was illustrated in Fig. 10. Klemp and Rotunno (1983) argue that this downdraft is typical for a supercell storm evolving into the tornadic stage and is distinct from the rear-flank downdraft. The occlusion downdraft is thought to be induced by a downwarddirected pressure gradient that responds to the sudden low-level buildup of vertical vorticity and associated pressure falls (Carbone 1983; Klemp and Rotunno 1983; Brandes 1984a; Hane and Ray 1984; Brandes et al. 1988). The present case study is the first to have sufficient resolution to accurately depict this downdraft and assess its forcing mechanism. Gal-Chen (1978) first proposed the use of syntheses from multi-doppler radar analyses to retrieve the perturbation pressure and density patterns using a least squares method. The retrieval treats the pressure and temperature as unknown variables in the anelastic momentum equations and solves the derived Poisson equation using the Doppler-derived wind patterns and the radar reflectivity. The latter field is used as an estimate of the mixing ratio in the vertical momentum equation. The thermodynamic retrieval scheme proposed by Roux (1985, 1988), Roux and Sun (1990), and Roux et al. (1993) was applied to several of the radar volumes shown in this paper in order to facilitate the interpretation of the wind fields. A momentum check was performed to assess the quality of the retrievals. The values ranged from , which is within an acceptable range defined by Gal-Chen and Kropfli (1984). Based on the uniform sampling in time, the tendency in the anelastic equation was included in the retrieval. Hane and Ray (1984) assumed steady state and Brandes (1984b) used a period of 19 min to estimate the time derivative in their pressure calculations. Results shown in Fig. 10 suggest that this is too coarse. For more information on the retrieval methodology, the reader is referred to Roux et al. (1993). The kinematic wind field at was chosen in order to assess the forcing mechanism of the occlusion downdraft. A pressure fall of 9 mb is shown near the circulation center of the mesocyclone at 600 m (Fig. 11). Bluestein (1983) states that 3-mb surface pressure falls are characteristic of mesocyclones although the distance of each surface observing station to the circulation center was usually unknown. Davies-Jones and Kessler (1974) have documented one case of a 34-mb pressure drop. The isobar pattern in relation to the wind field reveals that the flow is in cyclostrophic balance. The two ridges of pressure