The Evolution of Low-Level Rotation in the 29 May 1994 Newcastle Graham, Texas, Storm Complex during VORTEX

Transcription

1 JUNE 2001 ZIEGLER ET AL The Evolution of Low-Level Rotation in the 29 May 1994 Newcastle Graham, Texas, Storm Complex during VORTEX CONRAD L. ZIEGLER, ERIK N. RASMUSSEN, AND TOM R. SHEPHERD NOAA/National Severe Storms Laboratory, Norman, Oklahoma ANDREW I. WATSON NOAA/National Weather Service, Tallahassee, Florida JERRY M. STRAKA School of Meteorology, University of Oklahoma, Norman, Oklahoma (Manuscript received 7 March 2000, in final form 7 October 2000) ABSTRACT This paper reports the results of an analysis of airflow evolution in the tornadic Newcastle Graham, Texas, storm complex of 29 May A series of seven pseudo-dual-doppler analyses from 2242 to 2315 are performed from tail radar observations by the National Oceanic and Atmospheric Administration P-3 aircraft. Subjective analyses of quasi-horizontal single-doppler radar observations provide a detailed look at structure and evolution of the hook echo and the low-level Newcastle mesocyclone. Special emphasis is placed on the evolution of lowlevel [i.e., below 1 km above ground level (AGL)] rotation of the parent mesoscale circulation of the Newcastle tornado and the origins of mesoscale rotation preceding tornadogenesis. The structure and evolution of the Newcastle and Graham mesocyclones are compared and contrasted. The airborne Doppler analyses reveal that the tornadic Newcastle cell had supercell characteristics and that the Newcastle storm circulation could be classified as a mesocyclone based on commonly accepted criteria of circulation amplitude, spatial scale, and persistence. The Newcastle mesocyclone initially developed downward from midlevels (i.e., 2 5 km AGL), then transitioned into a subsequent period of rapid low-level stretching intensification and upward growth just prior to the development of an F3 tornado. Single-radar analysis reveals the stretching contraction and intensification of the Newcastle mesocyclone and an embedded tornado cyclone prior to and after tornadogenesis. In contrast, the nontornadic Graham mesocyclone ultimately became rainfilled and transitioned from moderate stretching growth to negative stretching after the development of a central downdraft in low levels, possibly contributing to tornadogenesis failure. Using a hybrid, two-supercell schematic diagram to depict the Newcastle Graham storm complex, it was concluded that the Newcastle tornado occurred at the traditionally accepted location of a supercell tornado at the point of the warm sector occlusion in the westernmost cell. Computed trajectories based on a Lagrangian solution of the vertical vorticity equation suggested that the midlevel Newcastle mesocyclone was formed by a sequence of tilting of ambient horizontal vorticity followed by stretching intensification in the rotating updrafts. The air parcels that entered the low-level Newcastle mesocyclone initially possessed vertical vorticity of order 10 3 s 1, which was subsequently concentrated by stretching upon entering the Newcastle updraft to form the low-level mesocyclone. Though the vorticity dynamical origin of the weak ambient rotation could not be identified, the spatial origins of low-level trajectories that entered the Newcastle mesocyclone were determined to be from a broad area of low-level rainy easterly outflow from the Graham storm. The present findings were compared and contrasted with results of an earlier study of the Newcastle storm. 1. Introduction Understanding the mechanisms responsible for the origins of low-level [i.e., below 1 km above ground level Corresponding author address: Dr. Conrad L. Ziegler, National Severe Storms Laboratory, Mesoscale Research/Applications Division, 1313 Halley Circle, Norman, OK Ziegler@nssl.noaa.gov (AGL)] vertical rotation in thunderstorms is fundamentally important to the development of improved severe storm forecasting and nowcasting methods. Of perhaps greatest significance is the ability to anticipate tornadogenesis and to discriminate those relatively few storms that will actually produce tornadoes from the larger body of severe storms. Yet, anticipating the development of the larger-scale parent low-level circulation of the tornado is also important, in that the tornado 2001 American Meteorological Society

2 1340 MONTHLY WEATHER REVIEW VOLUME 129 forecasting problem may be viewed as a series of conditional probabilities (e.g., probability of low-level rotation given severe convection, probability of tornadogenesis given low-level rotation, etc.). For example, the results of Trapp (1999) illustrate how only a fraction of low-level mesocyclones may produce tornadoes. 1 The significance of ambient low-level vertical rotation as a precursor for tornadogenesis is widely acknowledged (e.g., Burgess et al. 1993; Vasiloff 1993). It has previously been thought that a prerequisite for the generation of at least large tornadoes is the existence of a midlevel (i.e., 2 5 km AGL) mesocyclone (Davies- Jones 1985), a temporally persistent and spatially extensive region of intense vertical rotation within the main storm updraft (Burgess and Lemon 1990). However, only a fraction of (either low level or midlevel) mesocyclones produce tornadoes for reasons that are not clear (Burgess et al. 1979; Burgess and Lemon 1991). Previous studies suggest two categories of classical mesocyclones: those developing downward from midlevels (e.g., the Union City, Oklahoma, storm studied by Brown et al. 1978) and those developing upward from low levels (e.g., the Stillwater, Oklahoma, storm studied by Vasiloff 1993). A class of nonmesocyclonic, tornadic storms has been documented in which vertical rotation appears to originate in low levels, subsequently intensifying and extending vertically as a convective updraft develops above a pre-existing area of horizontal circulation in the boundary layer (Brady and Szoke 1989; Wakimoto and Wilson 1989; Roberts and Wilson 1995). A hybrid category of mesocyclone development has been reported by Wakimoto et al. (1998), in which a pre-existing boundary layer mesocyclone merges with a developing midlevel mesocyclone to form a deep circulation. The development of storm-scale rotation has been investigated in numerous studies (Klemp 1987). Rotunno (1981) and Rotunno and Klemp (1982) determined that the midlevel mesocyclone was created by the sequence of tilting of ambient horizontal vorticity as inflow entered the updraft, followed by additional growth from stretching as inflow accelerated upward. Davies- Jones (1984) showed that the tilting contribution to midlevel rotation in the storm was dependent on the existence of a storm-relative streamwise horizontal vorticity component in the inflow to the updraft. Regarding the initiation of low-level rotation, storm-scale solenoidal forcing in the boundary layer may amplify the component of the inflow s horizontal vorticity oriented parallel to the updraft gradient and thus provide a source of tilting generation (Rotunno and Klemp 1985). It has been argued that updrafts alone cannot be responsible for intense low-level rotation and tornadogenesis, since they would sweep newly generated vertical vorticity up- 1 WSR-88D radar climatological studies of the frequency of reported tornadogenesis in low-level mesocyclones are in progress (J. Trapp, NSSL, 2000 personal communication). ward toward middle levels (Davies-Jones 1982). For example, Davies-Jones and Brooks (1993) hypothesized that vertical rotation may be baroclinically forced and increased by tilting with descent in downdrafts, then further tilted and also intensified by stretching as air exits the downdraft and enters the base of the updraft. The goal of the present study is to document and gain insights into the origins of low-level mesoscale rotation preceding the Newcastle tornado of 29 May The Newcastle storm, one focus of intensive mobile mesoscale field observations during the Verification of the Origins of Rotation in Tornadoes Experiment (VOR- TEX; Rasmussen et al. 1994), was previously investigated by Wakimoto and Atkins (1996, hereafter referred to as WA) and Trapp (1999). The present study of the Newcastle storm contributes a pseudo-dual-doppler wind analysis, permitting a more quantitative interpretation of storm structure and evolution than WA, who employed a single-doppler radar analysis. An important and unsolved forecasting problem is the ability to anticipate the development of low-level rotation, requiring understanding of the differences between storms experiencing downward development of rotation from midlevels and storms whose low-level rotation originates within the boundary layer. Evidence will be presented that the Newcastle storm has some characteristics of both downward-developing midlevel and upward-developing low-level mesocyclones, and as such may be classified as a hybrid of the two former categories. The evolution of the Newcastle circulation is compared and contrasted with the evolution of the mesocyclone associated with the neighboring, nontornadic Graham, Texas, storm. In spite of their development in proximity to one another, thus implying near environments with similar severe weather potential, the Newcastle mesocyclone intensified sufficiently to produce an F3 tornado while the neighboring Graham mesocyclone did not intensify sufficiently for tornadogenesis to occur. The observations and analysis techniques used in the study are outlined in section 2, the analysis results are presented in section 3, a discussion and interpretation of the results are presented in section 4, and conclusions are presented in section Observations and analysis techniques a. NOAA P-3 aircraft observations The primary observational platform for this study was the National Oceanic and Atmospheric Administration (NOAA) P-3 aircraft, which provided airflow and reflectivity observations of storm structure, in situ measurements of the mesoscale storm environment, and video recording of the cloud field. The 5-cm lower fuselage (LF) radar provided radar reflectivity in quasi-horizontal 360 sweeps, while the 3-cm Doppler tail radar obtained radial velocity and reflectivity measurements within vertical (i.e., helical due to aircraft motion) sector sweeps.

3 JUNE 2001 ZIEGLER ET AL TABLE 1. Properties of the P-3 Doppler analysis legs. Determination of the nominal analysis times and the horizontal analysis offsets ( x/ y) are described in the text. Leg Beginning ending time (UTC) Nominal analysis time (UTC) Closest horizontal distance to Newcastle mesocyclone (km) x/ y offset (km) : : : : : : : : : : : : : : / 1.5 0/ 1.5 0/ 1.5 0/ / / / 1.6 Pseudo-dual-Doppler tail radar data were obtained by the alternative fore aft scan technique (AFAST; Jorgensen et al. 1996). The P-3 flew a series of seven consecutive legs along the southwest flank of the Newcastle storm from 2240 until 2318 (all times are UTC). The nominal analysis time for each leg was defined for the present study as the average of the times at which the fore and aft sweeps sampled the core of the developing Newcastle storm and its attendant circulation. A series of pseudo-dual- Doppler analyses (Table 1) were generated following methods discussed in appendix A. A photogrammetric analysis of airborne video and still images (appendix C) was used to infer the basic dimensions and location of the mature Newcastle tornado and the accompanying wall cloud. b. Generation of trajectories and Lagrangian analysis of vorticity dynamical evolution Air parcel trajectories were computed using trilinear spatial interpolation and a second-order Runge Kutta time integration scheme. The time integration scheme followed a two-step, predictor corrector approach: 1) compute parcel displacement over a time interval t from an initial point to the provisional Lagrangian point using the vector velocity interpolated to the initial point; 2) interpolate the vector velocity to the provisional Lagrangian point, average the initial and provisional velocity components, and determine a refined Lagrangian point by recomputing the parcel displacement from the initial point using the averaged velocity; and 3) repeat the above process for N steps, until either N t exceeds a time limit or the trajectory exits the analysis volume. The 3D fields of airflow and other derived quantities were assumed to vary linearly in time between the two input Doppler analyses closest to the current time of a Lagrangian point. Proceeding from the analysis of WA, who related stretching of pre-existing vertical vorticity along a mesoscale boundary to the intensification of the Newcastle circulation, we examined the vertical vorticity forcing by tilting and stretching. Departing somewhat from previous studies, we evaluated the latter forcing terms and the accumulated vorticity along Lagrangians approximating the motion of air approaching the developing circulation. The anelastic vertical vorticity equation took the form (Brandes 1984) d u w w u Fy Fx ( f ), dt z y z x x y x y tilting stretching diffusion (1) where is vertical vorticity, (u,, w) are the Cartesian wind components, f is the Coriolis parameter, and F is the frictional force. The only practical means of estimating the contribution of diffusion to vorticity evolution would have been to apply a first-order or K-theory closure, a fairly crude approximation of the turbulent stress based on the Doppler-derived airflow deformation. Brandes (1984) used a K-theory closure for mixing to estimate that the turbulence term was at least one order of magnitude smaller than stretching within supercell tornadic mesocyclones. Hence, the vertical vorticity equation in the present study was simplified by neglecting the turbulence term and evaluating only the tilting and stretching terms. While mixing effects could possibly be significant in the surface layer, due to possibly large airflow deformation and vertical turbulent fluxes, such surface layer effects are not considered here owing to the lack of airflow observations in the lowest m. The simplified vertical vorticity equation [i.e., Eq. (1) minus diffusion] was integrated in time following the motion as a first-order ordinary differential equation (i.e., model in subsequent discussion), producing a history of Lagrangian vertical vorticity evolution. The time integration of the vorticity equation was accomplished as part of the predictor corrector scheme used to compute the associated air trajectory. The fields of airflow, vorticity, and vorticity forcing by tilting and stretching were interpolated from the analysis grid to the Lagrangian point during integration. The initial condition for vertical vorticity was evaluated at time t 0, and integration proceeded forward to time t 1 (i.e., time limits defined by t 0 t t 1 ). To constrain a given trajectory to end at a specified point, a backward trajectory from time t 1 to time t 0 was initially computed from that point,

4 1342 MONTHLY WEATHER REVIEW VOLUME 129 followed by integration of the vorticity equation along the subsequent forward trajectory. Truncation error in highly curved flow was minimized by using an adequately short time step ( t 5 sec) to ensure that the backward and forward trajectory displacements essentially coincided. To assimilate observed data to the greatest extent possible, the stretching term was computed in the linearized form ( 0 f 0 )( u/ x / y), with a subscript zero indicating an observed value and the constant Coriolis parameter f 0 taking the value 10 4 s 1. Since integration time was explicitly limited by locations of data boundaries relative to a Lagrangian point, and was also implicitly limited by the accumulation of truncation and wind analysis bias errors with time, only those trajectories that ended in the Newcastle cell at 2253 and 2305 were integrated. To limit accumulated integration errors in the vorticity forcing terms, the Lagrangian vorticity evolution was calculated between and Two time-spaced analyses (2250/2253 and 2301/2305, respectively) were incorporated over the 6-min periods of the vorticity integration, thus defining the temporal and spatial gradients of velocity and vorticity forcing as parcels encountered strong shears and trajectory curvatures on entering the Newcastle cell. However, the positions of air trajectories were computed through the periods and The justification for computing air trajectory positions over longer time periods than the vorticity integrations is that the boundary layer flow was observed to change rather gradually during the 8-min time gaps (i.e., the periods and ). 3. Results a. Preconvective environment and storm initiation The Newcastle storm developed within a line of isolated, deep convection over north-central Texas during the midafternoon of 29 May In expectation of storm development, the NOAA P-3 research aircraft was flying over north-central Texas as early as 1900 and began flying legs adjacent to developing cells around Several Mobile Cross-chain Loran Atmospheric Sounding System (M-CLASS) soundings (Rust et al. 1990) were obtained in the preconvective and nearstorm environments. Convective initiation occurred along a convergence boundary that intersected an outflow boundary over north-central Texas (Figs. 1a,b,d). The combination of surface conditions (Fig. 1a), aircraft traverses (Fig. 1c), an aircraft descent sounding (Fig. 1e), and an M-CLASS sounding (Fig. 1f) suggest that the convergence boundary resembled a weak cold front due to the following characteristics: 1) the presence of a westward-tilted trough with height through a deep tropospheric layer above the boundary (WA); 2) gradually decreasing air and dewpoint temperatures with increasing distance to the west of the boundary; 3) a shallower boundary layer to the west of the convergence line than to its east; and 4) northwesterly flow and backing winds with height west of the boundary, suggesting regional-scale cold advection. On the other hand, evidence of downward mixing of warm dry air west of the boundary and the detection of a thermal maximum along the convergence line suggested a hybrid character including some similarity to a weak (i.e., small acrossboundary moisture difference), though distinctly nonclassical dryline. In contrast, the true dryline (e.g., Ziegler and Rasmussen 1998; Crawford and Bluestein 1997) was restricted to the Big Bend region of southwest Texas. Extreme convective instability and vertical wind shear within the warm sector southeast of the outflow trough merger point supported the forecast for severe thunderstorms given that deep convection had been initiated. Multilevel prestorm penetrations of the weak cold frontal boundary, including the low-level pass around 1935 (Figs. 1c,d), collectively revealed evidence of deep, sustained lifting of boundary layer moisture that was conductive to convective initiation (Ziegler and Rasmussen 1998). The 2047 National Severe Storms Laboratory (NSSL) NSSL-2 sounding (Fig. 2) revealed an environment with very high values of convective available potential energy (CAPE), rather low convective inhibition (CIN), and considerable vertical wind shear that could support the development of mesocyclonic updraft rotation. The early stage of mesoconvective development was dominated by the emergence of strong storms near Wichita Falls and Graham, Texas, by 2130, as other storms developed southwestward along the weak cold front (Figs. 3a,b). It is noted that the 2047 NSSL-2 sounding had sampled the air mass between the Wichita Falls and Graham storms. The storms rapidly intensified and moved southeastward between 2130 and 2245 (Figs. 3c,d), the latter time just after the first P-3 Doppler analysis to be reported. Apparent chaff echoes located to the west of the weak cold frontal boundary remained separated from the storms. The Newcastle cell and tornado were developing in the western portion of the analysis domain at 2315 (Fig. 3e). The Graham storm continued to move southeastward after 2315 (Figs. 3e,f). b. Storm evolution A southward-propagating, arc-shaped reflectivity band driven by outflow from the Graham storm preceded the development of the Newcastle cell (Figs. 4a,b). At 2242 (Fig. 4a) the arc-shaped reflectivity band was anchored by two counterrotating mesoscale vortices at 3 km, the Graham mesocyclone 2 to the east around the 2 Brandes (1984) used a threshold vertical vorticity value of s 1 to delineate mesocyclones in dual Doppler analyses of tornadic storms.

5 JUNE 2001 ZIEGLER ET AL FIG. 1. Atmospheric conditions on the afternoon of 29 May (a) Surface state at 2100 UTC, with solid-contoured mean sea level pressure (add 1000 to contour level for units of mb) and dashed-contoured dewpoint temperature ( C); (b) visible satellite imagery at 2100; (c) in situ P-3 observations at 150 m AGL relative to surface boundaries ( ); (d) view from P-3 looking east from location d in (c) at ; (e) P-3 descent sounding ending at 1930 at location P-3 in (b) and (c); (f) 1955 NSSL-4 sounding at location NS4 in (b) and (c). The station models in (a) and (c) include winds (half barb 2.5ms 1, full barb 5ms 1 ), temperature ( C) at upper left, dewpoint temperature ( C) at lower left, and MSL pressure (mb) at upper right in (a) only. The gray lines at point d in (c) denote the viewing limits of the P-3 nose video camera. The P-3 temperature in (c) was increased about 1.5 C to approximate the local surface temperature assuming a dry adiabatic lapse rate. A heavy solid curve with open triangles denotes a mesoscale cold front, a heavy scalloped curve denotes a dryline, and a heavy dashed curve denotes a convergence line. Symbols W, N, and G in (b) locate Wichita Falls, Newcastle, and Graham, TX, respectively. (x, y) coordinates (km) of (23, 12) and a mesoanticyclonic circulation around (16, 12). An inflow notch into the mesoanticyclone had formed around (15, 13) at 2242 and 3 km. With subsequent development (Figs. 4b,c), the Graham mesocyclone moved southward while the trailing, western edge of the arc-shaped reflectivity band moved slowly westward. By 2250 (Fig. 4b), the Newcastle cell had begun forming northwest of the inflow

6 1344 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 2. Skew T logp plot of 2047 NSSL-2 sounding on 29 May notch into the anticyclonic circulation. The precursor rotation of the midlevel Newcastle mesocyclone at this developing stage was evident as a small area of increasing midlevel vertical vorticity centered around (9, 14) (Fig. 4b), increasing to mesocyclone strength around the same location by 2253 (Fig. 4c). A low-level easterly outflow surge from the precipitation area on the north side of the Graham mesocyclone was evident at all analysis times. The outflow surge from the Graham storm assumed the character of a convergence band with enhanced lowlevel vertical vorticity immediately to its east by 2250 (i.e., shear convergence line, Fig. 4b). A low-level circulation approaching mesocyclone intensity that was restricted to the boundary layer (i.e., boundary layer mesocyclone 1) was noted at (10, 15) from 2250 to 2253 on the northwest end of the shear convergence line (Fig. 4b). A second boundary layer mesocyclone was located just east of the shear convergence line at (14, 14) and 2253 (Fig. 4c). It should be noted that the boundary layer mesocyclones were shallow and existed away from deep, moist convection, and therefore needed to be distinguished semantically from true mesocyclones. The Newcastle cell had developed a prominent hook echo and intense low-level and midlevel mesocyclones near (9, 12) by 2301 (Fig. 4d). Parker and Keighton (1994) reported that either three-dimensional correlated shears or mesocyclones were indicated in midlevel WSR-88D measurements during several volume scans between 2249 and Considering the proximity of the low-level Newcastle mesocyclone at 2301 to boundary layer mesocyclone 1 at 2253 and the general westward movement of the shear convergence line between 2253 and 2301, it is possible that the boundary layer mesocyclone 1 phased with the midlevel Newcastle mesocyclone. It is also possible that the boundary layer mesocyclone was purely incidental to the descent of the midlevel mesocyclone into the lowest 1 km and its subsequent intensification. Due to the absence of additional analyses and the large evolution over the 8-min period from 2253 to 2301, the action of the phasing process cannot be confirmed with the present data. This hypothetical phasing process of the pre-existing boundary layer mesocyclone with the midlevel Newcastle mesocyclone could conceivably have assisted the development of the low-level Newcastle mesocyclone. In a study of mesocyclone and tornado development in the Garden City, Kansas, storm during VORTEX-95, Wakimoto et al. (1998) concluded that a pre-existing boundary layer mesocyclone phased with the midlevel mesocyclone and main updraft to produce rapid intensification of the low-level mesocyclone and ultimately tornadogenesis. Applying space-to-time conversion and an estimated westward motion speed of roughly 3 m s 1 to the photogrammetric analysis and damage survey (WA), the onset of F0 wind damage at the eastern end of the Newcastle tornadic damage track was inferred to have occurred around The Newcastle mesocyclone continued to intensify after 2302, moving westward to around (8, 12) by 2305 (Fig. 4e). The Newcastle mesocyclone continued its westward motion and further intensified through the time of the first tornado photo from a ground team 3 taken at 2309 (Fig. 4f). During the development of the low-level mesocyclone between 2301 and 2305, 6 8 m s 1 updrafts were noted south of the mesocyclone on the inflow side of the center of circulation (Figs. 4d,e). The westward-moving shear convergence line from the Graham storm overtook the Newcastle mesocyclone by The overall evolution of the low-level Newcastle mesocyclone bears some similarities to the paradigm described by Brandes (1984) and Rotunno and Klemp (1985). Figure 20 of Brandes (1984) and Fig. 5 of Rotunno and Klemp (1985) were rotated clockwise by 90 to roughly visualize their results relative to the present northwesterly flow case (Fig. 4). The Newcastle cell contained a wrapping updraft band that pivoted in a counterclockwise sense around the location of maximum vertical vorticity, and a rear downdraft was noted to the north and west of the developing circulation (Figs. 4d f). Both Brandes (1984) and Rotunno and Klemp (1985) noted the expansion of the wrapping outflow boundary to the south and west of the developing mesocyclone in observed and simulated supercells (respectively), a feature also seen in the low-level Newcastle mesocyclone. The character of the low-level Graham mesocyclone appears dominated by the southwardsurging outflow boundary, including development of an elongated, trailing band of updraft and vertical vorticity with several local maxima, a morphology not well described by the Brandes (1984) and Rotunno and Klemp (1985) studies. Vertical cross sections through locations of maximum vertical vorticity reveal the spatial correspondence of updrafts, precipitation cores, and rotation in the Newcastle cell and its mesocyclone (Fig. 5). At 2242 (Fig. 5a, cross section a-a ), vertical vorticity greater than 5 3 The condensate funnel formed no more than 1 min prior to the first tornado photo at 2309 (G. Stumpf, NSSL, 2000 personal communication; G. Stumpf is not affiliated with VORTEX).

7 JUNE 2001 ZIEGLER ET AL FIG. 3. Deep convection over north-central Texas on 29 May 1994 from GOES satellite imagery and composite WSR-88D network radar reflectivities. (a) Visible imagery at 2131; (b) WSR-88D radar composite at 2130; (c) same as (b) but at 2215; (d) same as (b) but at 2245; (e) same as (b) but at 2315; (f) same as (b) but at The grayscaled label bar denotes reflectivity (dbz). The pseudo-dual-doppler analysis domain is located by the black box s 1 was located within and upstream from the updraft above 1.5 km and west of x 13 km, with weaker or negative vertical vorticity at lower levels. Vertical vorticity exceeded mesocyclone intensity by 2250 above 4 km (Fig. 5b), more than doubling midlevel rotation strength in the updraft between 2242 and Vertical vorticity in the Newcastle mesocyclone continued to increase through 2253 (Fig. 5c). By 2301 the most notable features were a secondary maximum of vertical vorticity near the ground, the persistent collocation of the mesovortex and updraft, and the development of a rear-flank downdraft (RFD) to the west of the mesovortex (Fig. 5d). As previously noted, the secondary low-level vertical vorticity maximum could have been related to the phasing of the pre-existing boundary layer mesocyclone 1 with the midlevel Newcastle mesocyclone, among other possibilities. The low-level mesocyclone continued to intensify with a developing

8 1346 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 4. P-3 pseudo-dual-doppler analyses on 29 May 1994 at (a) 2242; (b) 2250; (c) 2253; (d) 2301; (e) 2305; (f) Columns from left to right: i) velocity and reflectivity (heavy gray contours at 10-dBZ interval from 20 dbz, gray fill 40 dbz) at 0.5 km; ii) vertical velocity (heavy gray contours at 2 m s 1 interval, gray-filled updraft) and vertical vorticity ( 10 3 s 1 ) (thin black contour at s 1 interval) at 0.5 km; iii) and iv), as i) and ii) except at 3 km and with vertical velocity at 20 m s 1 interval from 10 m s 1. Velocities are relative to the assumed motion of the Newcastle mesocyclone (u cell 3.4, cell 0.9).

9 JUNE 2001 ZIEGLER ET AL FIG. 4.(Continued) Heavy dashed curves are convergence lines. Vertical cross sections and horizontal nested subgrids are located by thin black lines and dashed boxes, respectively, while the P-3 flight track is denoted by black lines at lower left. The symbols G and N locate the Graham and Newcastle mesocyclones respectively, while the symbols SCL, RFD, BLM #1, and BLM #2 denote, respectively, the shear convergence line, rear-flank downdraft, and boundary layer mesocyclones described in the text.

10 1348 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 4.(Continued)

11 JUNE 2001 ZIEGLER ET AL FIG. 4.(Continued)

12 1350 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 5. Vertical west-to-east analysis cross section through the Newcastle cell, located as in Fig. 4. Analyses are presented for the following nominal map times: (a) 2242 (a-a ); (b) 2250 (b-b ); (c) 2253 (c-c ); (d) 2301 (d-d ); (e) 2305 (e-e ); (f) 2312 (f-f ). Newcastle mesocyclonebounded weak-echo region (BWER) through 2305 (Fig. 5e) to 2312 (Fig. 5f), when a maximum vertical vorticity of s 1 was analyzed in the lowest 250 m near the location of the F3 tornado. The midlevel vertical vorticity maximum evolved into a core of rotation from the ground through 5 km after 2301 as the lowlevel maximum became dominant. The analysis at 2315 (not shown) revealed the same overall morphology of the Newcastle mesovortex and its parent updraft. Following the definition of supercells as persistent deep, moist convective updrafts that exhibit a well-defined midlevel mesocyclone along with attendant hook echoes and BWERs (WA), it was concluded that the Newcastle cell possessed supercell storm characteristics. The low-level mesocyclone of the Graham supercell storm achieved its maximum intensity at 2250 (Figs. 4b and 6), the time of its tornadogenesis failure according to the definition of Trapp (1999). The Graham mesocyclone at its failure stage was compared with the Newcastle mesocyclone at its tornadogenesis time of 2305 (Trapp 1999) in the hope that features possibly leading to tornadogenesis failure might be identified. Notable morphological similarities included collocation of the midlevel mesocyclones with the main updrafts and the presence of BWERs on the east (inflow) flanks of the main updrafts (cf. Figs. 6b and 5e). A narrow precipitation core and a rear flank downdraft were present to the west of the low-level Newcastle mesocyclone at 2305, assisting development of low-level convergence beneath the midlevel Newcastle mesocyclone. In contrast, at 2250 a relatively broad precipitation core (i.e., the arc-shaped reflectivity band in Fig. 4b) displaced the low-level convergence to the east of the Graham mesocyclone, while a downdraft was displaced to the west side of the precipitation core. The implied reduction of low-level convergence beneath the midlevel Graham mesocyclone may have inhibited the development of low-level rotation by reducing stretching growth of vertical vorticity, a hypothesis to be examined further in section 4. The mesoanticyclone at 2250 was concen-

13 JUNE 2001 ZIEGLER ET AL FIG. 5.(Continued) relative vector airflow, reflectivity, and vertical vorticity are as indicated as in Fig. 4. The symbol BLM #2 denotes the boundary layer mesocyclone described in the text. trated above 3 km and around (15, 11) on the north side of the arc-shaped reflectivity band between the Graham and Newcastle supercells (Fig. 6a). Certain aspects of mesocyclone evolution were emphasized in a three-dimensional, perspective rendering of storm structure (Fig. 7). The most noteworthy feature was the downward development of the Newcastle mesocyclone between 2250 and 2253 in proximity to the previously noted boundary layer mesocyclones 1 and 2 (Figs. 7a c). The Newcastle mesocyclone broadened and also became continuous in vertical extent between the 2253 and 2301 analyses, a substantial evolution that, as previously noted, occurred during a period of rather poor time resolution caused by the long flight legs needed to cover both mesocyclones. A columnar precipitation core remained collocated with the tubelike vorticity core of the Newcastle mesocyclone, which developed a progressively greater eastward tilt and became narrower with height between 2301 and 2315 (Figs. 7c f). It is hypothesized that the apparent narrowing of the Newcastle mesocyclone prior to 2315 was caused by the gradual convergent horizontal contraction of the circulation leading up to the tornadic stage. The Graham mesocyclone achieved its greatest horizontal extent between 2253 and 2301, subsequently moving out of the analysis domain but leaving shallower mesocyclonic circulations in its wake. Other low-level and midlevel circulations of mesocyclone strength were observed between 2250 and 2315, including the boundary layer mesocyclones 1 and 2 prior to c. Morphology and evolution of the Newcastle radar hook echo Measurements from the LF radar (Fig. 8) revealed greater detail of the evolving low-level Newcastle cell structure than was possible from the objectively analyzed tail radar data. A well-defined hook echo had developed by 2301 (Fig. 8a). The hook developed a more complex structure by 2311, including an anticyclonic hook (Figs. 8b d) and an echo-weak hole at the inferred tornado location (Fig. 8d). Tail radar measurements from a single beam at nearly horizontal incidence were subjectively analyzed to ex-

14 1352 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 6. Velocity, vertical vorticity, and reflectivity in the Graham cell at Airflow is relative to the assumed motion of the Graham mesocyclone (u cell 0, cell 5.1). (a) Horizontal analysis section at 2.5 km; (b) vertical west-to-east analysis cross section through the Graham mesocyclone, located as in (a). Airflow vectors, reflectivity, and vertical vorticity are indicated as in Fig. 4. The symbols G and A locate the Graham mesocyclone and the mesoanticyclone, respectively. amine other details of the hook echo region of the Newcastle mesocyclone. The tail-aft (TA) radar measurements along all quasi-horizontal rays (i.e., rotation angle 0 ) were mapped to the analysis grid at the flight altitude of 0.75 km. A subjective interpretation of these single-doppler fields was used to refine the estimated velocity shears and reflectivity structure of the hook echo. At 2301 (Figs. 9a,b), a radial velocity differential of 42 m s 1 over 3 km (i.e., solid line segment in Figs. 9a,b) was centered on the reflectivity gradient on the western inside margin of the hook echo. A radial velocity differential of 42 m s 1 over 1.8 km (i.e., dashed line segment in Figs. 9a,b) was centered at the same location as the larger-scale shear. This smaller-scale radial velocity differential implied the existence of a developing tornado cyclone that was embedded within the outer mesocyclonic circulation (e.g., Brown and Wood 1991). The tornado cyclone was of slightly larger horizontal scale than the tornado itself, and resembled a tornadic vortex signature (TVS) in the sense that strongly opposing radial velocities were being detected at closely spaced gates on adjacent radials (e.g., Brown et al. 1978). Note that both the mesocyclonic and tornado-cyclonic velocity couplets were rotated clockwise relative to a line normal to the individual rays, indicating that the vortices at both scales were strongly convergent at this time (Brown and Wood 1991). By 2305 (Figs. 9c,d), the (outer) mesocyclonic velocity differential had increased to 50 m s 1 over 2.3 km, while the (inner) tornado-cyclonic velocity differential had increased to 48 m s 1 over 1.3 km. Based on the orientation of the velocity couplets, the mesocyclone and tornado cyclone at 2305 were convergent as at However, the mesocyclonic convergence was weaker than at The center of velocity shear had wrapped cyclonically around to the southeast quadrant of the developing hook echo, which increasingly enveloped the circulation center. By 2312 (Figs. 10a,b), the weakly convergent tornado-cyclonic velocity differential was 49ms 1 over 1.3 km while the center of velocity shear was collocated with the echo-weak hole feature at the center of the hook echo. In contrast to the intensifying tornado cyclone, the mesocyclone at 2312 had become nondivergent and had a velocity differential of about 40

15 JUNE 2001 ZIEGLER ET AL FIG. 7. Surfaces of constant vertical vorticity and radar reflectivity in perspective view of the full radar analysis domain on 29 May (a) 2250; (b) 2253; (c) 2301; (d) 2305; (e) 2312; (f) The solid yellow surface is the s 1 vertical vorticity, while the transparent green surface is the 40-dBZ reflectivity. The symbols N and G denote the Newcastle and Graham mesocyclones, respectively, while the symbols BLM #1 and BLM #2 denote the boundary layer mesocyclones 1 and 2 described in the text. The view is looking north, with the north and east side planes of the analysis domain in blue. The brown surface is ground level. ms 1 over 2.5 km. Parker and Keighton (1994) reported the detection of a TVS by WSR-88D radar at 2312 and about 2 km. At 2315 (Figs. 10c,d), the tornado-cyclonic velocity differential was 47 m s 1 over 0.2 km (i.e., the local along-track gate spacing of successive sweeps at the center of circulation), while the mesocyclonic velocity differential was about 40 m s 1 over 1.5 km. The center of velocity shear at 2315 coincided with the echoweak hole feature as at The aforementioned sequence of subjective singleradar analyses from 2301 to 2315 collectively suggested an initially increasing mesocyclonic shear that combined with a sustained increase in peak angular momentum due to vortex contraction and the development of a TVSlike, tornado-cyclone-scale vortex within the low-level mesocyclone. It is noted that the implied convergent intensification of angular momentum in the tornado cyclone between 2312 and 2315 broadly corresponded with the narrowing of the surface damage track. Since the tornado was on the ground during this time period (Fig. 11) and was responsible for the surface damage, the narrowing damage track and convergent intensification of tornado-cyclonic shear were globally consistent with convergent intensification of the tornadic vortex itself. The detailed mesocyclonic structure as revealed by single-doppler data was interpreted relative to the visual tornado appearance (Fig. 11) and objectively analyzed mesocyclone structure (Fig. 12). An annular region of circulation had developed around the tornado in the lowest 1 2 km by 2312 (Figs. 12a,b), persisting through 2315 (not shown). The wall cloud of the Newcastle tornado was located within this annular rotation volume (Figs. 11a,b, Fig. 12b). It is also noted that an RFD (w 2ms 1 at 0.5 km) had developed 1 2 km west of the low-level Newcastle mesocyclone by 2301 (Fig. 4d) and persisted through 2305 (Fig. 4e). The RFD was located west of the tornado at 2312 (x 3.5, y 11, Fig. 4f; x 3kmtox 6 km, Fig. 6f). By 2312, the RFD-induced convergence band had wrapped around the low-level mesocyclone and tornado (Fig. 12a), forcing low-level updrafts outward with a broad region of updraft across the midlevel mesocyclone (Fig. 12b). The analyses of the low-level mesocyclone at 2312 and 2315 were consistent with the aforementioned results of Brandes (1984) and Rotunno and Klemp (1985), who noted the developing RFD and expansion of the wrapping outflow boundary to the south and west of the developing low-level mesocyclone of isolated supercell storms. Analyzed shears were compared with the observed peak along-track shear across the center of the mesocyclone based on the single (TA) radar subjective anal-

16 1354 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 8. History of LF radar reflectivity during the growth stage of the hook echo in the Newcastle mesocyclone. (a) 2301:09; (b) 2304:06; (c) 2305:08; (d) 2311:28. The thin dashed curve outlines the radar hook echo, while the letter T denotes the inferred tornado location. Gray range rings are at 10 km intervals. The grayscaled label bar denotes reflectivity (dbz). The heavy black dot denotes the P-3 location. The thin-dashed boxes denote the locations of single radar subjective analyses in Figs. 9 and 10. ysis. The theoretical filtering effect of objective analysis was expected to be significant, implying larger peak V (i.e., tangential velocity difference across the mesocyclone) and vertical vorticity values than inferred from the objectively analyzed P-3 Doppler radar data. Assuming cylindrical vortex symmetry (WA), the largest observed V of up to 50 m s 1 (Figs. 9 and 10) would correspond to an inferred observed vertical vorticity o V/L of around s 1. The analyzed vertical vorticity maximum, about s 1 (Fig. 6f), was about 50% lower than the largest inferred observed vertical vorticity estimate due to the filtering of the objective analysis procedure. 4. Discussion a. Evolution of the Newcastle and Graham mesocyclones As a preliminary step to examine morphological differences between the Newcastle and Graham mesocyclones, height profiles of vertical vorticity, reflectivity, vertical motion, and the stretching forcing of vertical vorticity through each mesocyclone were constructed following a modification of the approach used by WA. In the initial step, the mesocyclone center (x c, y c )at each level was estimated by averaging the locations of the center of closed circulation and the maximum ver-

17 JUNE 2001 ZIEGLER ET AL FIG. 9. Subjective analyses of TA radar reflectivity and ground-relative radial velocity at about 0 elevation ( 750 m AGL) during the growth stage of the hook echo in the Newcastle storm. (a) Radial velocity at 2301; (b) reflectivity at 2301; (c) radial velocity at 2305; (d) reflectivity at Gate values of radial velocity and reflectivity (gray) are plotted at 300-m intervals. The heavy dot locates the maximum along-track radial velocity shear, while the black line segment joins the radial velocity extrema across the mesocyclone. The black dashed line segment joins the radial velocity extrema across the tornado cyclone, and corresponding radial velocity differential ( V) and separation distance of peak values is indicated. The dark gray-filled area is the Newcastle tornado damage track. tical vorticity in the Newcastle cell-relative wind field. If a closed circulation center could not be identified, the mesocyclone center was assumed to coincide with the maximum vertical vorticity. In the second step, the maximum vertical vorticity and its horizontal grid indices (i max, j max ) were identified at each level within a circular search area centered on (x c, y c ), the search radii for the Newcastle and Graham mesocyclones being 1.5 and 2.5 km, respectively. Filtering the i max (z) and j max (z) arrays and rounding filtered indices to the nearest integer value eliminated discontinuous changes of horizontal location caused by levels with very broad vorticity cores. The vorticity, average reflectivity, and vertical velocity profiles were obtained from the gridpoint values at (i max, j max ) filt, while profiles of vorticity stretching were computed from centered horizontal differences about that location. The vorticity and vorticity stretching values in the Newcastle cell were fairly weak in the lowest 1 km at 2242 (Fig. 13). By 2250, vorticity and stretching had increased in the lowest 1 km as the midlevel mesocyclone intensified and descended. However, the low-level and midlevel mesocyclone remained distinct at 2250 and Strong stretching production of vertical vorticity had developed around 4 km by 2250, subsequently descending toward the surface between 2250 and 2301 in response to the strengthening of low-level updrafts during this period. The increased stretching production in

18 1356 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 10. Same as in Fig. 9, but depicting subjective analyses of TA radar radial velocity and reflectivity during the tornadic stage of the Newcastle storm. (a) Radial velocity at 2312; (b) reflectivity at 2312; (c) radial velocity at 2315; (d) reflectivity at The width of the heavy dot coincides with the separation of the radial velocity couplet of the tornado cyclone at the lowest 1 km between 2253 and 2301 coincided with a rapid increase of low-level vertical vorticity to over s 1. The development of the intense lowlevel mesocyclone at 2301 was synchronized with the collision of the westward-moving shear convergence line with the Newcastle updraft (e.g., Figs. 4d and 6c) and the onset of observed surface wind damage from the Newcastle circulation as reported by WA. As previously discussed, the data collectively suggested the possibility that the pre-existing boundary layer mesocyclone 1 merged with the descending midlevel mesocyclone to assist the subsequent convergent intensification of the low-level mesocyclone. Due to the large spacing of the 2253 and 2301 analyses, during which period any mesocyclonic merger would likely have taken place, the alternative possibility that the boundary layer mesocyclone 1 was purely incidental to low-level mesocyclogenesis could not be ruled out. Vertical vorticity intensified further by stretching and deepened upward from ground level through the lowest 5 km of the Newcastle mesocyclone between 2301 and the time of tornadogenesis (Fig. 13). Strong stretching production of vertical vorticity generally pervaded the lowest 5 km of the mesocyclone during the period of reported surface damage through the time of tornadogenesis. Reflectivity increased during the early mesocyclonic growth stage, then weakened slightly and subsequently increased again, reflecting the evolution of the radar hook (e.g., Figs. 4d f) and the episodic expansion of the radar vault around 2305 (e.g., Figs. 6d f). In contrast to the Newcastle mesocyclone, the Gra-

19 JUNE 2001 ZIEGLER ET AL FIG. 11. Visual appearance of the Newcastle tornado and associated cloud dimensions. (a) Tornado photograph taken by lead author (CLZ) from P-3 at ; (b) summary of photogrammetric analysis of P-3 side-looking video relative to tornado photo (a); (c) photograph of the Newcastle tornado at approximately 2314 looking south-southwest (courtesy M. Biddle). ham mesocyclone underwent a more gradual intensification through the lowest 5 km prior to 2301 when the largest overall midlevel vertical vorticity values were analyzed (Fig. 14). A secondary relative maximum of low-level vertical vorticity was evident prior to As previously noted, a broad precipitation core had begun spreading through the low-level Graham mesocyclone by 2250 (Fig. 6b), the inferred time of tornadogenesis failure. Stretching production of vertical vorticity was moderately strong at all levels prior to 2253, followed by the development of a central downdraft in the lowest 3 km leading to negative stretching below 2 km by The subsequent development of a central downdraft coincided with increasing reflectivities as precipitation spread through the entire low-level circulation (e.g., Figs. 4d,e). Due to the large spacing of the 2253 and 2301 analyses, as previously noted, the onset time of the low-level central downdraft could only be approximated. The most important difference between the evolution of the low-level mesocyclones was the development of persistent, strong positive stretching growth of vertical vorticity in the Newcastle mesovortex and development of negative vorticity stretching in the Graham mesovortex. b. Evolution of low-level rotation: Organizing stage of Newcastle mesocyclone (2253) The vorticity evolution of airflow entering the base of the Newcastle updraft at 2253 was examined with a series of 11-min integrations of the Lagrangian model described in section 2b. As previously noted, the vertical vorticity tendency was integrated over the shorter period of Only one sample trajectory has been shown for brevity, since several neighboring trajectories had similar characteristics. The Newcastle updraft was centered near (9, 15) and was undergoing rapid downward development and strengthening by 2253 (Figs. 4c and 13). A typical lowlevel trajectory ascended in the updraft to around (9, 13.5, 2.5) after a period of westward displacement (Fig. 15). The initial vertical vorticity had a value of order 10 3 s 1, reflecting ambient (i.e., pre-existing ) magnitudes of vertical vorticity in the boundary layer around the initial point. The pre-existing vertical vorticity was subsequently enhanced along the low-level trajectory by tilting and stretching in a mesoscale boundary layer updraft (Figs. 15b,d), followed by westward transport of this enhanced vorticity. The tilting generation occurred as the parcel experienced a positive crosswise horizontal vorticity component (i.e., horizontal vorticity component directed to the left of the flow) while traversing the northern flank of the mesoscale boundary layer updraft near (11, 15) (Fig. 15b). The observed parcel maximum vertical vorticity of s 1 at 2249 (Fig. 15d) represented the closest approach of the parcel to the boundary layer mesocyclone 1 (Fig. 4b), which was located near (10, 15). Due to the overall agreement of the modeled and observed vertical vorticity (Fig. 15d), it is concluded that the boundary layer mesocyclone 1 was formed by a combination of tilting and stretching in the boundary layer updraft. After a brief period of negative tilting and decreasing vertical vorticity, subsequent stretching and positive tilting increased vertical vorticity to its overall maximum value as it entered the Newcastle updraft and midlevel mesocyclone (Figs. 15c,d). The vertical vorticity of the inflow to the developing

20 1358 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 12. Velocity, reflectivity, and vertical vorticity structure of the Newcastle mesocyclone, depicted as in Fig. 6, and corresponding to tornado photographs in Fig. 11. (a) Horizontal section at 2312 and 0.5 km; (b) vertical north south cross section at 2312, located as in (a). Velocities are relative to the Newcastle mesocyclone. The thin dashed box in (b) outlines the approximate area above ground as viewed in Fig. 11a, while the thin black curves and dark gray fill denote the vertically scaled wall cloud and tornado dimensions depicted schematically in Fig. 11b. Velocities in (b) are plotted at every grid point. Thick dashed curves denote convergence lines. The dark gray filled area in (a) is the Newcastle tornado damage track. The gray line at lower left of (a) is the P-3 flight track, while the orientation of the side-looking camera view is denoted by the thin dashed black line. The location of the photo in Fig. 11a is denoted by P in (a). midlevel Newcastle mesocyclone was continuously swept upward by the updraft (e.g., Fig. 15), implying that some source of vertical rotation at very low levels was needed to develop the mesocyclone near the ground (e.g., Davies-Jones 1982). Following WA, it was hypothesized (see section 3b) that a phasing of the Newcastle updraft with a pre-existing low-level circulation such as boundary layer mesocyclone 1 could possibly have assisted the low-level mesocyclogenesis process. c. Evolution of low-level rotation: Mature stage of Newcastle mesocyclone (2305) The vorticity intensification and the spatial pattern of trajectories culminating in the low-level Newcastle mesocyclone at 2305 were examined via a series of 15- min integrations of the Lagrangian model (Fig. 16). The trajectories arrived at 45 intervals around the circumference of a circle at 0.5-km altitude and 1-km radius centered on the low-level mesocyclone. One strongly

21 JUNE 2001 ZIEGLER ET AL FIG. 13. Time height profile plots of selected radar analysis fields in the Newcastle mesocyclone from 2242 to Top panel: vertical vorticity (black contour) and reflectivity (gray contour); bottom panel: stretching production of vertical vorticity (black contour) and vertical velocity (gray contour). Gray and black bands along horizontal axis denote periods of surface damage only and existence of a visible tornado funnel, respectively. forced trajectory was subjectively picked to illustrate the vorticity forcing and net vorticity changes following the motion of air reaching the mesocyclone. As described in section 2b, the vertical vorticity tendency was integrated over the shorter period of Trajectories converged at 2305 through a roughly 90 sector into the low-level mesocyclone (Fig. 16a), reflecting the strong convergence at this time (i.e., strongly increasing updraft with height in low levels in Fig. 10). All parcels experienced a period of strong stretching within 5 min of reaching the mesocyclone (Fig. 16a). Several parcels moved northwestward just east (i.e., to the rear) of the Graham outflow shear convergence band, which extended into the Newcastle mesocyclone at this time. Trajectories began at altitudes between 80 and 340 m AGL (Fig. 16a) and remained in the range m AGL over the period from 2250 to Vertical vorticity growth from a point (b) southeast of the lowlevel mesocyclone was dominated by stretching from ambient values of order 10 3 s 1 (Fig. 16b). Since the objectively analyzed vertical shears and horizontal gradients of vertical velocity were rather small in the lowest several hundred meters, tilting did not contribute significantly to vertical vorticity growth along the low-level trajectories feeding the low-level mesocyclone at FIG. 14. Same as in Fig. 13, but for the Graham mesocyclone from 2242 to d. Origins of low-level rotation The air trajectory calculations generally suggested that parcels feeding the low-level Newcastle mesocyclone originated to its east within a broad area of convective outflow and rainfall in the vicinity of the Graham cell (Fig. 17). Two source regions of air ultimately reached the low-level Newcastle mesocyclone: 1) the cyclonically curved easterly flow north through east of the Graham mesocyclone (Figs. 17a,b) and 2) outflow from the convective line extending northwestward out of the arc-shaped reflectivity band (lower left corner of Fig. 17b). As will be elaborated on in subsequent discussion, the shear convergence line preceding the easterly outflow surge from the Graham storm could have contributed some vertical vorticity for subsequent stretching intensification in the Newcastle mesocyclone. However, since the mesocyclonic inflow trajectories originated from within an area much broader than the mesoscale convergence band and boundary layer mesocyclone 1, it was impossible to explain the origins of the low-level Newcastle mesocyclone entirely as convergent concentration of horizontal circulation along the pre-existing boundary. The outflow air parcels that subsequently fed the Newcastle mesocyclone had vertical vorticities of order 10 3 s 1, thus maintaining an influx of weak circulation about a horizontal circuit (i.e., horizontal circulation ). Since the low-level mesocyclone

22 1360 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 15. Storm structure relative to an air trajectory that entered the Newcastle mesocyclone at (a) Newcastle mesocyclone-relative velocity, reflectivity (heavy gray contours in increments of 10 dbz starting at 20 dbz, and gray fill 40 dbz), vertical vorticity ( 10 3 s 1 ) (thin black contour at s 1 interval), and path of trajectory d at 2253 and 0.5 km; (b) horizontal vorticity, vertical velocity (heavy gray contours in increments of 2 m s 1 and gray fill 2 ms 1 ), and tilting production of vertical vorticity (thin black contours in increments of s 2 ) at 2250 and 0.5 km; (c) velocity, reflectivity, and vertical vorticity at 2253 in vertical west east cross section j-j, located as in (a); (d) time series of stretching and tilting production of vertical vorticity (expressed as over t 5 s) and modeled and observed vertical vorticity along the 6 min trajectory that was initiated at the location denoted by the circled-x symbol in (a), (b), and (c). The heavy black vector in (b) denotes the interpolated horizontal vorticity at the parcel location at The symbols BLM #1 and BLM #2 denote the boundary layer mesocyclones described in the text. formed between 2253 and 2301 and since the tornado dissipated around 2329 (WA), an influx of horizontal circulation would need to have been sustained for at least 30 min to enable the development and maintenance of the Newcastle mesocyclone and tornado by stretching. It was concluded by WA that a low-level shear feature was stretched by the Newcastle updraft to produce the F3 tornado. This shear feature was seen in their and radial velocity analyses at 0.6 km (Figs. 14a,b, WA). The equivalent low-level shear feature in the present objectively analyzed tailforward (TF) radial velocities was located around (10, 14) at 2242 and An intrusion of strong easterly winds into the north side of the radar echo band west of the Graham storm was noted around (10, 14) and 0.5 km in Figs. 4a,b. The vertical vorticity of this area of easterly winds was dominated by positive y component shear (i.e., / y s 1 ), explaining the shear feature noted by WA. The spatial coincidence of the

23 JUNE 2001 ZIEGLER ET AL FIG. 16. Air trajectories arriving at the low-level Newcastle mesocyclone at 2305 and 0.5 km. (a) Horizontal trajectory locations and initial altitudes (km) relative to Newcastle mesocyclone-relative velocity, reflectivity (heavy gray contours in increments of 10 dbz starting at 20 dbz, and gray fill 40 dbz), and stretching production of vertical vorticity (thin black contours in increments of s 2 ) at 0.5 km; (b) time series of stretching and tilting production of vertical vorticity (expressed as over t 5 s) and modeled and observed vertical vorticity along the 6-min trajectory that was initiated at the location denoted by the circled-x symbol in (a). Arrowheads denoting endpoints of trajectories are excluded for clarity.

24 1362 MONTHLY WEATHER REVIEW VOLUME 129 FIG. 17. Origin locations (denoted by circled- mark) of air trajectories arriving at the Newcastle mesocyclone. (a) Origin point at 2242 of 11-min trajectory arriving at mesocyclone at 2253; (b) origin points at 2250 of 15-min trajectories arriving at mesocyclone at The panels denote Newcastle mesocyclone-relative velocity vectors, reflectivity (heavy gray contours in increments of 10 dbz starting at 20 dbz, and gray fill 40 dbz), and vertical vorticity ( 10 3 s 1 ) (thin black contour at s 1 interval) at 0.25 km. Initial trajectory altitude (km) is indicated next to each point. shear feature noted by WA and the present boundary layer mesocyclone 1 suggest that they represented the same circulation. The aforementioned trajectory calculations thus implied that the shear feature was maintained by a combination of tilting and stretching in the mesoscale boundary layer updraft. The trajectory calculations also suggested that vertical vorticity generated by the mesoscale boundary layer updraft was subsequently advected into the midlevel mesocyclone, thereby playing a direct role in the mesocyclogenesis process even though the boundary layer mesocyclone 1 and the Newcastle mesocyclone were not collocated at It is unlikely that a small circulation such as the boundary layer mesocyclone 1 could have sustained convergent vortex intensification for a period of up to 30 min, though it could have contributed to the early low-level mesocyclogenesis. The vorticity analysis thus implies that the low-level rotation was supplied by processes external to the low-level boundary layer mesocyclone or a pre-existing boundary, but what was the ultimate origin of this rotation? With the exception of the 2253 trajectory, which revealed a contribution to mesocyclonic rotation via tilting in a mesoscale boundary layer updraft, none of the other trajectories that subsequently reached the low-level mesocyclone revealed a contribution from tilting to initiate vertical rotation. Instead, the common feature of the trajectories was the inflow of rather low ambient values of vertical vorticity and their subsequent rapid intensification by stretching in the main updraft, as also found by Wakimoto et al. (1998). Descent to the left of the outflow axis of the mesoscale downdraft within the Graham outflow could conceivably also have generated vertical vorticity by a combined baroclinic-tilting process (Davies- Jones 1996), though present data were insufficient to evaluate this hypothesis. It is believed that the combination of rather poor spatial sampling and resolution and the correspondingly heavy analysis filtering may have mitigated against computing significant values of low-level tilting production (Wakimoto et al. 1998). e. Schematic diagram of the Newcastle Graham storm complex The observations summarized herein were used to develop a schematic diagram of the Newcastle Graham storm complex at its mature stage. The foundation of the Newcastle Graham schematic was the conceptual isolated supercell schematic developed by Lemon and Doswell (1979). The Lemon Doswell supercell schematic was rotated through 90 to account for the northwesterly flow in the present case. The storm complex was conceptualized by representing the Newcastle and Graham storms as adjacent supercells organized in a line (e.g., U.S. Dept. of Commerce 1990), along with inclusion of the arc-shaped reflectivity band structure and (elevated) mesoanticyclonic cell seen in the radar analysis. The two supercell mesocyclones were thus spatially separated by the elevated mesoanticyclonic updraft. The juxtaposition of two supercells oriented with the winds aloft required a merger of the forward flank rainy downdraft of the upwind (Newcastle) cell with the rear flank downdraft of the downwind (Graham) cell. It was assumed that the analyzed shear convergence bands corresponded to the cold outflow boundaries noted in the schematic diagram. In this hybrid form of the Lemon Doswell schematic (Fig. 18), supercell tornadoes would be probable at the two ends of the warm sector occlusions. Additionally, gust front tornadoes could occur on the wrapping outflow boundaries southwest of the supercell tornado locations. Based on ground visual surveillance of the Graham and Newcastle cells by the various storm intercept teams, tornadoes did not develop at either of the gust

25 JUNE 2001 ZIEGLER ET AL FIG. 18. Conceptual model of the mature Newcastle Graham storm complex in the lowest 1 km, as inferred from the Doppler analyses and derived from classical conceptual models described in the text. Heavy solid curves are mesoscale cold fronts, heavy dashed contour denotes the precipitation shield, thin black arrows are airflow streamlines, and light and dark shading denote updraft and downdrafts areas, respectively. The circled T symbols indicate possible tornado locations. front tornado locations or the downwind supercell tornado location. Based on the pseudo-dual-doppler analyses presented herein, it has been concluded that reinforced convergence along and to the rear of the forward flank baroclinic zone provided the additional stretching to amplify the upwind parent low-level mesocyclone of the upwind (Newcastle) cell. The enhanced low-level stretching hypothetically also supported tornadogenesis. The parent updraft of the Newcastle tornado had classical supercell characteristics (e.g., persistence, strong midlevel and low-level rotation, and a wrapping gust front or convergence boundary). Although the period of time that the Newcastle updraft had existed prior to mesocyclogenesis ( 10 min) was apparently short in comparison to typical supercell storm lifetimes of several hours, the Newcastle storm nevertheless had adequate time to develop a distinct supercell morphology. f. Findings of the present study relative to Wakimoto and Atkins (1996) As the present analysis proceeds from the previous study of the Newcastle storm (WA), it is appropriate to compare the results and conclusions of the two studies. The present study suggests some similarities, as well as some differences, with respect to both the morphological interpretations and the origins of low-level rotation preceding tornadogenesis in comparison to the findings of WA. The following salient issues are considered. 1) The present study confirms the conclusions of WA that a tornado was not produced by the neighboring Graham supercell mesocyclone and that the Newcastle tornado was associated with a separate convective cell. 2) The present analysis indicates that the spatial dimensions and timescale of the Newcastle storm circulation were consistent with its classification as a mesocyclone, a critical question in view of the differing conceptual models associated with mesocyclonic versus nonmesocyclonic tornadoes. The results of the present study indicate that a midlevel updraft and mesocyclone preceded the Newcastle tornado, while WA concluded that a midlevel mesocyclone was not present prior to the tornado. However, WA referred to the Newcastle circulation as a mesocyclone during its most intense phase. The present study notes that the Newcastle cell had supercell storm characteristics using the definition provided by WA. Hence, WA classified the Newcastle tornado as nonsupercellular while the present study reveals new evidence that the Newcastle tornado had supercell characteristics. Some differences in nomenclature and interpretation between the two studies could have arisen from the amount and nature of the synthesized data (e.g., single vs pseudo dual Doppler) and the specific procedures employed in the respective objective analyses. Other differences in interpretation may have arisen from the conceptual models used to assist in interpreting the observations. 3) Assuming a classical isolated supercell conceptual model, WA concluded that the Newcastle tornado developed at the conceptual location of a gust front tornado relative to the parent (Graham) mesocyclone. In contrast, the present results imply that the Newcastle and Graham cells were too widely separated to apply the one-supercell conceptual model. More importantly, the present results indicate that the Newcastle storm possessed supercell characteristics. Therefore, the present analysis is more consistent with a two-supercell conceptual model of the Newcastle Graham storm complex, wherein the Newcastle tornado was associated with the conceptual supercell tornado location in the westernmost (upwind) cell. 4) The present results imply that stretching predominates in the initiation of the low-level Newcastle mesocyclone, partially confirming the hypothesis advanced by WA. However, the results of the two studies differ in their interpretation regarding the origins of the rotation that is subsequently stretched. Further, WA concluded that the low-level mesocyclone and the tornado itself formed from the stretching of vertical vorticity of a pre-existing shear feature or boundary. In contrast, the present trajectory analysis implies that air entering the low-level mesocyclone originated in outflow from the Graham storm and neighboring convection to the east of any localized outflow boundaries or boundary layer circulations.