Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL and BLUESTEIN 2563 Hand, the Source of Low-Level Rotation Is More Contro- Versial

2562 MONTHLY WEATHER REVIEW VOLUME 125

The Arcadia, Oklahoma, Storm of 17 May 1981: Analysis of a Supercell during Tornadogenesis

DAVID C. DOWELL AND HOWARD B. BLUESTEIN School of Meteorology, University of Oklahoma, Norman, Oklahoma (Manuscript received 29 October 1996, in ®nal form 10 February 1997)

ABSTRACT On 17 May 1981, an extensive dataset was collected for a supercell thunderstorm that produced an F2 tornado near Arcadia in central Oklahoma. Coordinated dual-Doppler scans of the storm by 10-cm research radars were collected at approximately 5-min intervals from 30 min before the tornado touched down until 15 min after the tornado had dissipated. The Arcadia storm was also well sampled by a 444-m-tall instrumented tower. The low- level in¯ow, updraft, mesocyclone, and rear precipitation core of the supercell all passed across the tower. A comparison of the instrumented tower measurements with a dual-Doppler synthesis reveals that the latter qualitatively resolved the low-level ¯ow. However, the magnitudes of the low-level horizontal winds and updraft speed were underestimated. In addition, the vertical shear of the horizontal wind in the lowest kilometer was unresolved in the Doppler winds. In the storm environment, horizontal vorticity was strong (ϳ1.5 ϫ 10Ϫ2 sϪ1) and approximately streamwise over the depth of the instrumented tower. Just upstream (northeast) of the updraft, the magnitude of horizontal vorticity was nearly twice this value and had likely been enhanced by baroclinic generation of horizontal vorticity and/or stretching of horizontal vorticity. Tilting of the resulting horizontal vorticity into the vertical produced the pretornadic low-level mesocyclone. Low-level mesocyclone in¯ow was primarily from the east, but during the tornadic stage, parcels approaching from the north and west were also drawn into the circulation. The tornado formed southeast of the mesocyclone center and near the tip of the re¯ectivity hook echo while low-level mesocyclone vorticity was increasing. Tornadogenesis occurred near the nose of the rear downdraft within a region of horizontal shear between southeasterly in¯ow into the storm and westerly out¯ow from the rear downdraft. Pressure retrievals suggest the rear downdraft south of the mesocyclone center was associated with a downward-directed perturbation pressure gradient force. The tornado and the parent storm dissipated as out¯ow surged eastward ahead of the updraft. This case study is the ®rst to include a comparison of independent measurements of the wind ®eld in and near the low-level mesocyclone of a supercell. The wind analysis is also complemented by the instrumented tower thermodynamic measurements.

1. Introduction TEX) in the southern and central plains of the United States (Rasmussen et al. 1994). Radiosondes, mobile Recent progress in numerical simulations and ad- mesonet instrument packages, portable Doppler radar, vances in observational capabilities have sparked re- airborne Doppler radar, and other implements were all newed interest in veri®cation of the details of how a directed toward the broad goal of understanding the for- supercell thunderstorm (Browning 1964) produces a tor- mation of tornadoes. nado. Computer simulations of supercells are becoming Tornadogenesis relies on horizontal convergence more and more successful at resolving tornadolike vor- within the boundary layer to amplify vertical vorticity tices (Wicker and Wilhelmson 1995; Grasso and Cotton to magnitudes characteristic of tornadoes (Ward 1972; 1995). With the nearly nationwide coverage provided Lewellen 1993). In some cases, thunderstorms concen- by the WSR-88D network, tornadic storms are being trate vertical vorticity already present in the storm en- sampled relatively frequently by Doppler radar (Guer- vironment (Wilson 1986; Brady and Szoke 1989; Wak- rero and Read 1993; Magsig and Burgess 1996). During imoto and Wilson 1989). In contrast, supercell thun- 1994±95, scientists collaborated on the Veri®cation of derstorms appear to act upon low-level vertical vorticity the Origins of Rotation in Tornadoes Experiment (VOR- produced by the storm itself (Barnes 1970; Brandes 1984b; Rotunno and Klemp 1985). Tilting in the storm updraft of horizontal vorticity associated with the environmental vertical wind shear Corresponding author address: David C. Dowell, School of Me- teorology, University of Oklahoma, 100 East Boyd, Room 1310, Nor- is generally believed to be the source of rotation for the man, OK 73019-0628. midlevel mesocyclone of a supercell (Barnes 1970; E-mail: [email protected] Brandes 1984b; Rotunno and Klemp 1985). On the other

᭧1997 American Meteorological Society

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2563 hand, the source of low-level rotation is more contro- versial. Numerical simulations by Rotunno and Klemp (1985) suggest that the development of low-level rotation awaits the presence of evaporatively cooled air near the surface. Once the storm's cold pool is estab- lished, horizontal vorticity produced baroclinically along the cool air boundary upstream of the updraft is tilted into the vertical, and the vertical vorticity is ampli®ed by stretching within the updraft. In contrast to the work of Rotunno and Klemp (1985), the simulations by Walko (1993) demonstrate that environmental horizontal vorticity can be the source for low-level rotation. (Although environmental horizontal vorticity may itself have been generated by baroclinic effects, it is distinguished from the type of baroclinically generated horizontal vorticity described by Rotunno and Klemp by its existence prior to storm formation.) Wicker (1996) stresses the importance of phasing of horizontal vorticity of both types in the development of the low- level mesocyclone. An issue that remains even more ambiguous is how the low-level mesocyclone and tornado are related. Tor- nadic vortex signatures (TVSs) in Doppler radar data FIG. 1. Map of the locations of the instrumented tower, sounding (Brown et al. 1978) provide observational evidence of (Edmond), radar sites (Norman and Cimarron), and town of Arcadia, the vastly different magnitudes of size and vorticity in Oklahoma. The Norman radar is located at the origin. The tornado mesocyclones and tornadoes. Furthermore, the behavior occurred south of Arcadia between approximately 1700 and 1710 CST 17 May 1981. Contours of the re¯ectivity factor (dBZ) measured of the TVS within the mesocyclone varies from storm by the Cimarron radar at 1.0 km AGL at 1704 CST are also shown. to storm; it may appear ®rst aloft, ®rst at low levels, or The crescent-shaped region indicates where the dual-Doppler be- simultaneously over a large depth (Brown et al. 1978; tween-beam angle is greater than 45Њ (and less than 135Њ). Trapp and Mitchell 1995). The diagnosis of how mesocyclones and tornadoes are related continues to stretch the limits of observational capabilities. Another issue Oklahoma City. A high-resolution cross section of the that remains unclear is how much the processes that wind and temperature structure both within the pretor- produce low-level rotation vary from case to case; the nadic storm (including the low-level updraft and me- variety of observed storm types (Doswell and Burgess socyclone) and its environment were obtained, affording 1993) hints that not all tornadogenesis mechanisms are a unique opportunity for intercomparison of multiplat- alike. Investigation of such issues was a major moti- form observations. The tower measurements document vation for VORTEX (Rasmussen et al. 1994). details of the boundary layer ¯ow that are relevant to In this paper, we examine a dataset that has awaited the development of low-level rotation but that cannot analysis many years but that relates directly to the goals be resolved by distant Doppler radars. of VORTEX. On 17 May 1981, an extensive dual-Dopp- The main purposes of the research described in this ler dataset of an isolated, tornadic supercell was col- paper are twofold: to compare independent measure- lected with the National Severe Storms Laboratory ments of the wind ®eld within the 17 May 1981 su- (NSSL) 10-cm research radars (Fig. 1). Dual-Doppler percell and to look for clues about the processes of low- volumes span a period beginning 30 min before an F2 level mesocyclogenesis and tornadogenesis. tornado formed near Arcadia, Oklahoma, and ending 15 min after the tornado had dissipated. Previous studies of tornadic supercells using ground-based multiple- 2. Description of the 17 May 1981 dataset Doppler radar have not bene®ted from such a complete a. The Arcadia storm and its environment coverage of the supercell life cycle surrounding and including the tornadic stage (Brandes 1978, 1981; Ray Conditions on 17 May 1981 were characteristic of a et al. 1981; Brandes et al. 1988). In addition, the mean classic tornado outbreak in Oklahoma (Taylor 1982). A interval between consecutive dual-Doppler volumes (5 short-wave trough at 500 mb over the southern Rockies min) affords better time resolution than is typically during the morning moved into the Plains during the achieved in airborne Doppler radar studies (Dowell et day. Ahead of the trough, a warm front at the surface al. 1997; Wakimoto et al. 1996). raced northward through Oklahoma, and a dryline ad- A 444-m-tall instrumented tower sampled the Arcadia vanced eastward into west-central Oklahoma. Afternoon supercell as the storm moved over the north side of temperatures near 30ЊC and dewpoints over 20ЊC yield-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2564 MONTHLY WEATHER REVIEW VOLUME 125 ed a potentially unstable environment over central Okla- homa. Thunderstorm development began near the intersec- tion of the warm front and dryline in northern Oklahoma around 1400 CST (all times in CST); later convective initiation occurred progressively farther south along the dryline (Brewster 1984). First echoes of what was to become the Arcadia storm appeared on the NSSL Nor- man radar display before 1500. A sounding taken from Edmond, Oklahoma (less than 10 km west of where the tornado later occurred), at 1430 (Figs. 1 and 2) indicates that the environment was characterized by substantial conditional instability and strong vertical shear of the horizontal wind. Since in the environment the convective available potential energy (CAPE) was 2250 J kgϪ1 (2500 J kgϪ1 if computed using virtual temperature), the 0±6-km shear near 5 ϫ 10Ϫ3 sϪ1, and the bulk Richardson number 22, the environment was supportive of supercell formation (Weis- man and Klemp 1984). Despite the strong vertical wind shear, the vertically integrated storm-relative environmental helicity from 0 to 3 km AGL was only 110 m2 sϪ2, that is, below the threshold value for tornadoes of 160 m2 sϪ2 suggested by the work of Davies-Jones et al. (1990). Owing to counterclockwise curvature in the hodograph between 2 and 4 km AGL (Fig. 2b), the 2±3-km layer contributes no net helicity. Davies and Johns (1993) have suggested that the height of the level of free convection (LFC) (nearly 2 km AGL in this case) (Fig. 2a) may be a more appropriate storm in¯ow depth to use in the helicity integration. The 110 m2 sϪ2 helicity from 0 to 2 km for the Edmond wind pro®le has the same helicity density as a 0±3-km layer with the threshold 160 m2 sϪ2 helicity. Between approximately 1700 and 1710 on 17 May 1981, an F2 tornado carved a path 5-km long south of the town of Arcadia, Oklahoma (Taylor 1982), destroy- ing a mobile home, downing power poles, and causing heavy damage to timber (D. Burgess 1993, personal communication). Twelve dual-Doppler volumes of the Arcadia storm were collected by the Norman and Ci- marron 10-cm research Doppler radars in central Okla- homa (Fig. 1) over a period of nearly an hour beginning at 1628. Thus, a relatively detailed documentation is available of the storm before, during, and after the tornado, except for a 10-min gap between dual-Doppler volumes during the pretornadic stage, when the Norman FIG. 2. Sounding taken from Edmond, Oklahoma, at 1430 CST 17 radar was brie¯y focused on another supercell in the May 1981. (a) Skew T±logp diagram. The solid and dashed lines vicinity. indicate temperature (ЊC) and dewpoint (ЊC), respectively. Full (half) wind barbs represent 5 m sϪ1 (2.5 m sϪ1); ¯ags represent 25 m sϪ1. Throughout the period covered by the Doppler data- Winds are shown at heights indicated in km AGL. Tower observations set, the geometry was favorable for dual-Doppler anal- at 444 m AGL at 1638 CST (within updraft) are plotted as closed ysis; the between-beam angle exceeded 45Њ for the dots. (b) Hodograph. Each circle represents5msϪ1 of wind speed. mesocyclone region of the storm (Fig. 1). During the The heights are in kilometers AGL. The small triangle indicates ob- Ϫ1 Ϫ1 tornadic stage, the mesocyclone was approximately 45 served storm motion (u ϭ 10ms ,␷ϭ6ms ). km from each radar; at this range, the low-elevation beams were 400±500 m above the ground, which was below cloud base. When Doppler data collection ended

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2565 at 1725, the Arcadia storm was losing supercell char- cess typically yielded nonzero vertical velocities at the acteristics. lower boundary. Therefore, a constant horizontal divergence correction was added to each column that would, upon reintegration of the continuity equation, b. Dual-Doppler analysis methodology require the vertical velocities to be zero at both the The Norman (Cimarron) radar collected data at 1.0Њ ground and the storm top; this procedure is commonly (0.6Њ) azimuthal increments; typical elevation angle in- termed the linear O'Brien (1970) adjustment. crements between sweeps were 0.8Њ±2.0Њ. Each radar Since the objective analysis scheme produced a qual- has a beamwidth of 0.8Њ. Re¯ectivity and radial velocity itatively smooth horizontal wind ®eld, no further data from each radar were interpolated to Cartesian grids smoothing was applied to the horizontal velocities. Ver- with horizontal and vertical grid spacings of 0.8 km and tical velocities were noisier in appearance; a single-pass 0.5 km, respectively. Given the 1.0Њ azimuthal incre- Leise (1981) ®lter was applied in order to eliminate ment between beams for the Norman radar, the 0.8-km features with wavelengths shorter than 2 km. Quantities spacing corresponds to the distance between adjacent such as vorticity were computed using centered ®nite gates during the tornadic stage of the supercell, when differences. Doviak et al. (1976) and Ray et al. (1980) the storm was 45 km away from the radar. The vertical provide a discussion of the errors involved in dual- grid structure was chosen such that the lowest level of Doppler wind synthesis. Mesocyclone-scale (several ki- the grid was at ground level (a physical boundary), and lometers), but not tornado-scale (several hundred me- the next lowest level was near the height of the low- ters), features will be resolved in the dual-Doppler anal- elevation radar sweeps. A spherical in¯uence region of yses. radius 1.2 km was used for the Cressman (1959) inter- polation to the grid. The radius of in¯uence was large c. Characteristics of the instrumented tower enough to account for the vertical spacing between consecutive sweeps and to produce a relatively smooth wind The low-level in¯ow region, mesocyclone, updraft, ®eld. Objectively analyzed data at ground level were and rear precipitation core of the pretornadic Arcadia extrapolated downward from the height of the lowest storm all passed across the 444-m-tall instrumented tow- sweeps (400±500 m AGL). Thus, implicit in the Dopp- er (Carter 1970) on the north side of Oklahoma City ler analysis is the assumption that velocities (and di- (Fig. 1). Horizontal wind, vertical velocity, temperature, vergence) are constant in the lowest 500 m. wet-bulb temperature, and other variables were mea- Ground clutter, range-folding contamination, sidelobe sured at 1.5-s intervals at up to seven levels between contamination, and other noisy data were removed be- the surface and 444 m. In the analysis, instrumented fore the analysis. Range-folding contamination was sig- tower data were smoothed using a 20-s running mean. ni®cant for this dataset owing to distant severe storms Since pressure measurements made on the tower appear north of the Norman radar; however, this contamination to be erroneous, they were not used in the analysis. primarily affected the light precipitation region in the central and northeast portions of the storm. Radial ve- 3. Storm morphology and evolution locities were dealiased manually. We used an empirical relationship between the radar re¯ectivity factor and The Arcadia storm was a mature supercell by the time raindrop terminal fall speed to subtract the component it was observed simultaneously by the Norman and Ci- of the latter from the interpolated radial velocities. marron radars, as evidenced by a well-de®ned meso- To account for motion of the storm while data were cyclone±mesoanticyclone couplet in the early scans at being collected, the data were advected horizontally to midlevels (Fig. 3a). (For the purposes of this paper, we locations corresponding to a central time of each data refer to a relatively long-lived storm with updraft ro- volume. The beginning of each volume was 1±2 min tation on scales of several kilometers as a ``supercell.'') before the center time. Storm motion was estimated by The storm had already produced, and was continuing visually correlating re¯ectivity features on the south- to produce, golfball-sized hail (Taylor 1982). For the west side of the storm between successive volumes. The horizontal scales resolved in the analysis, mesoanticy- analyses con®rmed that this storm motion estimate clone vorticity at 5.0 km AGL tended to exceed meso- matched that of the low-level mesocyclone during the cyclone vorticity in magnitude (e.g., Ϫ0.019 sϪ1 vs period surrounding and including the tornado. ϩ0.012 sϪ1 at 1647 CST) during the pretornadic stage The three-dimensional wind ®eld was obtained from (Fig. 3b). the dual-equation system using iterative downward in- Trajectories (not shown) are consistent with vortex tegration of the anelastic mass continuity equation sub- couplet production by the tilting of horizontal vorticity ject to the condition w ϭ 0 at the storm top; downward associated with environmental vertical shear. The air on integration was chosen because it tends to damp the the northwest side of the mesoanticyclone and the far vertical propagation of errors during the integration south side of the mesocyclone originated at midlevels, (Ray et al. 1980). Due to inexact upper boundary con- while the air within the southeast part of the meso- ditions and accumulation of divergence errors, this pro- anticyclone and within much of the mesocyclone as-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2566 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 3. Cimarron radar re¯ectivity factor (dBZ) and horizontal storm-relative winds at 5.0 km AGL. The x (east±west) and y (north± south) distances (km) are relative to the Norman radar. Local maxima in cyclonic and anticyclonic vertical vorticity are marked ``C'' and ``A,'' respectively. Updrafts are hatched. (a) 1634 CST, 0.5 h before tornado, w Ͼ 15 m sϪ1 shaded; (b) 1647 CST, 0.25 h before tornado, w Ͼ 20 m sϪ1 shaded; and (c) 1722 CST, just after tornado, w Ͼ 11 msϪ1 shaded.

cended from low levels. This ®nding agrees with con- ®guration like that observed in the pretornadic stage of ventional theory, which states that midlevel mesovor- the storm (Figs. 3a,b). tices in supercells arise from tilting of environmental During the time period roughly coinciding with the horizontal vorticity by the primary storm updraft tornadic stage, the midlevel vortex couplet gradually (Barnes 1970; Brandes 1984b; Rotunno and Klemp shifted from a northwest-to-southeast orientation (Fig. 1985). Horizontal vortex tubes (normal to the vertical 3b) to a northeast-to-southwest con®guration (Fig. 3c). shear) associated with the mean shear between the sur- This may be related to weakening updrafts at 5 km at face and 4 km AGL (Fig. 2b) were oriented from south- this time (Fig. 4). As the updraft dwindled, the meso- east to northwest; tilting of such vortex tubes by an anticyclone was advected downstream to the northeast updraft would produce a mesocyclone±anticyclone con- (Fig. 3c) relative to the mesocyclone; in contrast, the

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2567

weakening, multicellular, disorganized structure to the storm. The most notable change in the low-level pretornadic horizontal winds between 1630 and 1643 was the development of strong storm-relative northerlies on the southwest side of the storm, which occurred coincident with a growing appendage in re¯ectivity (Fig. 5). At 1630 there were strong in¯ow southerlies (and southeasterlies) along the entire southern ¯ank of the storm. At 1643 air still entered from the south to the east of the mesocyclone, but the wind ®eld southwest of the mesocyclone was dominated by a surge of out¯ow northerlies. By 1651, the storm-relative low-level winds on the southwest side of the mesocyclone had backed to north- westerly behind the advancing gust front (Fig. 5c). Then, with the development of westerlies south of the vorticity maximum, the low-level storm-relative ¯ow became circular (Figs. 5d,e). The onset of such storm- relative westerlies within the hook often signi®es the transition to the tornadic stage (Brandes 1978, 1981). A sharp increase in Cimarron radar Doppler velocities FIG. 4. The maximum vertical velocity vs height within the updraft Ϫ1 of the Arcadia storm for various times. (up to 30 m s ground relative) in the southern (outbound) member of the inbound±outbound couplet immediately preceded tornado touchdown. The tornado eventually dissipated as the out¯ow surged rapidly east- mesocyclone tended to remain collocated with the upward (Figs. 5e,f), leaving the low-level mesocyclone draft (e.g., Fig. 3a), even while the updraft was weak- behind. ening. Investigation of the magnitudes of the terms in Immediately before the tornado touched down, a pro- the vorticity equation for the mesoanticyclone (not nounced re¯ectivity hook developed on the southwest shown) con®rms that the relative importance of hori- side of the isolated supercell (Fig. 6). The Arcadia tor- zontal advection was increasing compared to the other nado was associated with a local horizontal shear anom- terms (stretching, tilting, and vertical advection) as the aly in the single-Doppler radial velocities (Brown et al. midlevel updraft was weakening. 1978) at the tip of the hook (Brandes 1981). This feature Close inspection of the supercell storm reveals that shows up most clearly in the low-level Cimarron radial it consisted of multiple cells (Dowell et al. 1997) during velocities at 1704 and the low-level Norman Doppler the Doppler radar observation period. The lobe of rel- velocities at 1710 (not shown). No clearly de®ned shear atively high re¯ectivity at 1634 CST around x ϭϪ10, anomaly was present prior to the tornadic stage. In these y ϭ 45 is associated with both a weak vorticity couplet respects, this case is similar to the 20 May 1977 Del and an updraft (Fig. 3a). This small cell moved rapidly City storm (Brandes 1981). We will refer to the shear northeastward away from the primary cell. It appears anomaly as a TVS, although it lacks the time continuity that storm splitting (Klemp and Wilhelmson 1978) was of a classic TVS (Brown et al. 1978). Gate-to-gate in progress when Doppler scanning began. The primary shears at 1704 (not shown) were 4.0 to 6.8 ϫ 10Ϫ2 sϪ1 cell had much stronger updrafts (Fig. 3a, around x ϭ in the lowest six elevation scans by the Cimarron radar, Ϫ6, y ϭ 37) and higher re¯ectivities than the smaller which is on par with what was observed in the Del City left-split cell; the primary cell eventually became tor- storm (Brandes 1981). nadic. Tornadogenesis coincided with the time when the While the tornado was forming, a third cell was de- storm top was collapsing (Fig. 7), re¯ectivity was gen- veloping on the south ¯ank of the storm east±southeast erally decreasing, and midlevel updrafts were weak- of the tornadic cell. By 10 min after tornado dissipation, ening (Fig. 4) (Fujita et al. 1976; Burgess 1982). How- re¯ectivity in the posttornadic core had decreased to ever, the dual-Doppler measurements suggest the low- less than 50 dBZ (around x ϭ 17, y ϭ 55 in Fig. 3c) level updrafts were stronger during the tornadic and im- and had reached 60 dBZ in the new core (around x ϭ mediate pretornadic stages (Fig. 4). Tornadogenesis also 25, y ϭ 50 in Fig. 3c). This new cell may have been corresponded to the time when low-level mesocyclone invigorated by the low-level convergence associated vorticity was increasing rapidly (Fig. 8). At 1643, mid- with the surging out¯ow (Fig. 5e) from the tornadic level vorticity exceeded that at low levels; during tor- cell. However, the new cell failed to become tornadic nadogenesis (1651 to 1704) midlevel vorticity was in- (Taylor 1982). Later single-Doppler scans indicated a creasing in magnitude, but low-level vorticity was in-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2568 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 5. Cimarron radar re¯ectivity factor (dBZ) and horizontal storm-relative winds at 0.5 km AGL; x and y as in Fig. 3. Gust front positions are marked by a dashed line: (a) 1630 CST; (b) 1643 CST; (c) 1651 CST; (d) 1704 CST, TVS location denoted by a dot and damage path indicated by a heavy line; (e) 1710 CST, TVS location denoted by a dot and damage path indicated by a heavy line; and (f) 1717 CST. creasing even more rapidly, as is common in tornadic The tornado formed within the mesocyclone in a zone supercells (Brandes 1984a). After the tornado had dis- of strong vertical velocity gradient (but within the up- sipated, the vertical gradient of vorticity was reversed; draft) near the nose of the rear downdraft. Further details low-level vorticity exceeded midlevel vorticity after of tornadogenesis will be discussed in section 5. 1715, as is also commonly observed (Brandes 1984a). To see if any of the ¯ow features close in size to the The vertical velocity ®eld at 1704 at 1.0 km AGL minimum resolvable wavelength (ϳ2 km) could be bet- (Fig. 9) exhibits classic tornadic structure (Lemon and ter illustrated, we reanalyzed the Doppler volumes at Doswell 1979; Brandes 1984b); an arc-shaped updraft 1704 and 1710 on a ®ner grid (500-m horizontal grid region surrounds a downdraft at the rear of the storm. spacing with a 500-m radius of in¯uence). With such a

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2569

FIG.5.(Continued) small radius of in¯uence, numerous gaps between data sampled directly by the tower (Fig. 11b). Both indicate in the vertical were produced in the grid, which pre- strong east±southeasterly in¯ow into the storm and cluded the computation of vertical velocities on the ®ne strong northerlies within the growing re¯ectivity ap- grid. However, since the radial velocities at elevation pendage on the southwest side of the storm. angles near zero are relatively unaffected by vertical Although the independent measurements of wind di- particle motion, we are able to estimate well the hori- rections within the storm agree, the dual-Doppler wind zontal velocities at low grid levels where there are radial speeds tend to be underestimates of the tower wind velocity data from both radars. The resulting analyses speeds (Fig. 11c). Some of the differences in wind speed show more detail in the ¯ow (Fig. 10). may be the result of storm evolution. At the time of the The TVS at 0.5 km AGL was within 500 m of the dual-Doppler analysis (1638), the strength of the storm- estimated damage path of the tornado (Fig. 10). The relative in¯ow was decreasing, while the strength of the tornado formed southeast of the center of the low-level storm-relative out¯ow was increasing (Figs. 5a,b). At mesocyclone and was associated with a local vorticity the edges of Fig. 11c, where corresponding measure- maximum within the larger-scale cyclonic ¯ow. ments were made approximately 10 min apart, it would thus appear to be improper to compare the values. For the measurements made within 4 min of 1638, 4. Instrumented tower data the mean difference in magnitude between dual-Doppler The low-level mesocyclone and updraft of the Ar- and tower winds is 6 m sϪ1. Sources of disagreement cadia storm passed the instrumented tower on the north may include (a) slightly different heights of the mea- side of Oklahoma City (Fig. 1) around 1638. We will surements (444 m for the tower, 500 m for the Doppler compare the dual-Doppler analysis centered at this time analysis); (b) differences in measurement type (point vs to the tower data through the use of time-to-space con- volumetric samples); (c) differences in ®ltering of the version of the latter (Barnes 1973; Goff 1976; Johnson raw data; and (d) ground clutter contamination of the et al. 1987). For a storm that is relatively steady state, Doppler velocities. the time cross section through the storm at low levels A comparison of the vertical shear of the horizontal can be thought of as a spatial cross section. A detailed wind measured by the tower between the ``surface'' (7 quantitative comparison of tower versus Doppler wind m AGL) and 444 m AGL with that in the Doppler anal- measurements is possible but is beyond the scope of ysis over the 500±1000-m layer revealed little corre- this paper. Instead, we will focus on a few key simi- lation between the two (Fig. 12). Differences in the larities and differences in the analyses of the low-level southwestern portion of the storm may be partly affected ¯ow. by the heights over which the shear is calculated, rel- Storm-relative dual-Doppler winds at 1638 at 500 m ative to the depth of the storm out¯ow. However, this AGL (Fig. 11a) capture many of the features in the ¯ow may not explain all dissimilarities within the storm.

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2570 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 6. Re¯ectivity factor scan by the Cimarron radar at 0.7Њ at 1658 CST. Range is given in kilometers.

Strong vertical shear (1.5 to 2.5 ϫ 10Ϫ2 sϪ1 over the indicate a strong updraft within the cyclonically curved depth of the tower) was typical of the storm environ- low-level ¯ow (Figs. 11a, 13). However, the analyzed ment, including the in¯ow portion of the storm (Fig. updraft centers are displaced 1 km from each other, 12b). The environmental sounding (cf. Fig. 2) indicates which results in a slight difference in the analyses of strong shear (approaching 1 ϫ 10Ϫ2 sϪ1) also over the the phase relationship between the updraft and the re- 500±1000-m layer. However, the Doppler analysis region of maximum cyclonic curvature in the horizontal solves very little shear in the storm in¯ow (Fig. 12a). winds. The displacement may be the result of a south- We believe this highlights a limitation of the Doppler westward tilt with height of the low-level updraft (Fig. data; horizontal vorticity associated with the vertical 14); the Doppler radars did not sample the vertical vari- shear of the horizontal wind is not resolved at low levels. ation in convergence in the lowest 500 m necessary to We conclude that a detailed vorticity budget associated resolve the updraft slope. with tornadogenesis near the surface is not possible. The maximum vertical velocity measured by the tow- Lack of resolution at low levels has been a problem in er within the updraft is roughly a linear function of general for multiple-Doppler datasets of tornadic storms height in the lowest 266 m (not shown). However, above (Brandes 1984b, 1988). 266 m the linear relationship breaks down; the maxi- Both the in situ and remotely sensed measurements mum updraft speed at 444 m is only slightly greater

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2571

FIG. 7. Time series of the height of the Arcadia storm top, as indicated by the 40-dBZ contour in raw Cimarron radar scans.

FIG. 8. The maximum vertical vorticity (10Ϫ2 sϪ1) vs height within than at 266 m. Since the convergence was mostly con- the mesocyclone of the Arcadia storm for various times. centrated below the height of the lowest Doppler-radar beams, it is not surprising that the peak updraft speed (Fig. 13) in the Doppler analysis (nearly 3 m sϪ1)is The retrievals around 1638 (not shown) indicate a only half that measured by the tower (6 m sϪ1). The slight downward-directed perturbation pressure gradient Doppler data are limited by the radar horizon and by force within the low-level (1.0 km AGL) updraft, which smoothing inherent in both the remote data collection is inconsistent with the observation that nonbuoyant air and the postanalysis. Overall, the comparison of the was being lifted in the updraft. Furthermore, the thermo- dual-Doppler analysis and instrumented tower data sug- dynamic retrieval (not shown) failed to resolve the ob- gests that the former is able to resolve features in the low-level winds with horizontal scales of around 2 km or longer but that the magnitude of the wind speeds in such features may be underestimated. The tower observations and the Edmond sounding (Fig. 2a) indicate that the updraft air at 444 m is neither saturated nor buoyant; a virtual temperature de®cit in the updraft of approximately1Kissuggested relative to the environmental air sampled by the instrumented tower 1 h previously (and by the Edmond sounding 2 h previously). The rising motion must have been associated with an upward-directed perturbation pressure gradient force (Brandes 1984a; Wicker and Wilhelmson 1995). With additional lift of around 1 km without en- trainment, the air would have become buoyant (Fig. 2a). We applied a pressure retrieval to the Doppler analyses to determine whether an upward pressure gradient force could be resolved within the low-level updraft. Perturbation pressure and virtual temperature were retrieved using the technique of Hane and Ray (1985), except that the velocity time tendency terms were also incorporated into the current study. Each retrieval is based upon a pair of consecutive dual-Doppler volumes. Velocity time derivatives (estimated by ®nite differ- FIG. 9. Horizontal storm-relative winds and vertical velocity (m ences) and mean velocities are computed in a storm- sϪ1) at 1.0 km AGL at 1704 CST; x and y as in Fig. 3. Contours for relative reference frame. negative values are dashed. The TVS location is marked with a dot.

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2572 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 10. Horizontal storm-relative winds and vertical vorticity (10Ϫ2 sϪ1) at 0.5 km AGL for the high-resolution (500-m grid spacing) analyses; x and y as in Fig. 3. Contours for negative values are dashed. The locations of the TVS (closed dot) and tornado damage path (thin solid line segment) are marked: (a) 1704 CST and (b) 1710 CST. served temperature gradient of 3 K kmϪ1 across the 1643 in Fig. 14a) is associated with high dewpoints mesocyclone (Fig. 14b). The radar horizon problem, within the heavy rain and area of strongest descending ground clutter problems, and smoothing in the Doppler motion. analysis limit the ability of the thermodynamic retrieval to resolve the low-level temperature ®eld. The thermodynamic retrieval was further hindered by the pres- 5. Low-level mesocyclogenesis and tornadogenesis ence of hail in the storm. Since the Arcadia storm was a. Theory from numerical simulations and laboratory a proli®c producer of golfball-sized hail (Taylor 1982) experiments and contained radar re¯ectivity factors as high as 70 dBZ, there is considerable uncertainty in the retrieval The frictionless Boussinesq vorticity equations are owing to dif®culty in relating the re¯ectivity factor to given by wץ precipitation mixing ratio. Additionally, there may be D␨ (w ϩ ␨ ; (1 ١´error owing to the general dif®culties in applying a ϭ ␻hh zץ thermodynamic retrieval to a volume with incomplete Dt data (Hane and Ray 1985). D␻ (ϫ (Bk), (2 ١ v ϩ(١´ h ϭ (␻ A vertical cross section within the low-level meso- Dt hh cyclone (Figs. 14a,b) reveals two features that are rem- iniscent of a density current (Goff 1976): a sloped baro- where D/Dt represents the total derivative, ␨ is the ver- clinic zone and a relatively narrow updraft that precedes tical vorticity, ␻ is the 3D vorticity, h subscripts denote it. However, the horizontal winds indicate that no gust horizontal components, B is buoyancy, and the other front surge along the direction of storm motion (east- quantities have their usual de®nitions. The ®rst term on northeast) has yet developed. At the time of this cross the right-hand side of (1) represents tilting of horizontal section (1635±1645), the out¯ow was surging south- vorticity into the vertical; the second is the stretching ward in a storm-relative sense (Fig. 5b). Shear in the of vertical vorticity. The ®rst term on the right-hand wind component normal to the cross section is the dom- side of (2) includes stretching of horizontal vorticity inant feature in the horizontal winds (Fig. 14c). and tilting of vortex lines; the second represents baro-

Equivalent potential temperature ␪e in the in¯ow and clinic generation of horizontal vorticity. updraft air was as high as 349 K (Fig. 14a). In the rainy Signi®cant vertical vorticity is already present in downdraft air on the west side of the mesocyclone, ␪e some thunderstorm environments and can be readily am- was as low as 335 K; the Edmond sounding (Fig. 15) pli®ed by stretching at the base of a thunderstorm up- suggests that this air had descended from above 750 mb. draft (Wilson 1986; Brady and Szoke 1989; Wakimoto

The secondary ␪e maximum farther west (t ϭ 1642 to and Wilson 1989). We will focus here on how meso-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2573

FIG. 11. Low-level horizontal storm-relative winds and Cimarron radar re¯ectivity factor (dBZ) at 1638 CST at 0.5 km AGL; x and y as in Fig. 3. (a) Dual-Doppler analysis winds at 0.5 km AGL at 1638 CST. (b) Time-to-space conversion of winds measured by the instrumented tower at 444 m AGL over a 30-min period centered at 1638 CST. (c) Vector difference between the winds in (a) and those in (b).

cyclogenesis and tornadogenesis can occur when such mechanism is insuf®cient to explain tornado-intensity vorticity is not already available. rotation (i.e., vorticity greater than 1 sϪ1) in the lowest According to conventional theory, supercell thunder- few hundred meters of the atmosphere (Davies-Jones storms derive their rotation (about a vertical axis) from and Brooks 1993; Walko 1993). By the time tilting and vorticity that is initially purely horizontal (Barnes 1970; stretching in the updraft have given an air parcel tor- Brandes 1984b; Rotunno and Klemp 1985). Tilting of nadic vertical vorticity, the parcel has risen far away horizontal vorticity into the vertical by the primary from the surface. Instead, Walko (1993) and Davies- storm updraft, followed by stretching of vertical vor- Jones and Brooks (1993) hypothesize that a downdraft ticity in the updraft, is the traditional conceptual model is also required in the genesis of the tornado's parent for the formation of a mesocyclone (vertical vorticity circulation. տ 0.01 sϪ1). However, recent work suggests that this An idealized numerical experiment by Walko (1993)

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2574 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 12. Vertical shear of the low-level horizontal winds and Cimarron radar re¯ectivity factor (dBZ) at 1638 CST at 0.5 km AGL; x and y as in Fig. 3. (a) Vertical shear in the 500±1000-m AGL layer in the dual-Doppler analysis winds at 1638 CST. (b) Time-to-space conversion of vertical shear between 7 and 444 m AGL measured by the instrumented tower over a 30-min period centered at 1638 CST.

demonstrates how a downdraft aids in low-level vortexgenesis. This process is investigated in terms of the circulation (Rotunno and Klemp 1985) about a material curve (i.e., the area integral of the component of vorticity normal to a surface outlined by the curve) in the ¯uid. Circulation is de®ned by

C ϭ Ͷ v´d l, (3)

where v is the 3D velocity, l is the space coordinate, and the integration is taken along a closed curve. The circulation about a curve moving with the ¯uid changes only owing to baroclinic and friction effects. In the Walko (1993) experiment, the circulation of a low-level vortex is derived from the environmental vertical shear of the horizontal wind; a simpli®ed model of how this mechanism works is depicted in Fig. 16. There is a circulation about the initially tilted material curve owing to the vertical wind shear. This circulation is conserved as the downdraft tips the material circuit FIG. 13. Vertical velocities measured by the instrumented tower at into a horizontal plane just above the surface. Equiva- 444 m AGL (heavy line) and synthesized in the dual-Doppler analysis lently, weak vertical vorticity is produced via tilting of at 0.5 km AGL at 1638 CST (contours, m sϪ1); x and y as in Fig. 3. Contours for negative values are dashed. The time-to-space conver- horizontal vorticity over a region many radii larger than sion of tower data covers the period within 15 min of the center time the eventual vortex core (Walko 1993). When the ma- (1638 CST). The heavy straight line indicates the direction of storm terial circuit then enters the base of the updraft, the low- motion. The displacement of the curve above (updraft) or below level vortex spins up as the curve converges. The sit- (downdraft) this line indicates the vertical velocity. The maximum and minimum vertical velocities are 6.1 m sϪ1 and Ϫ6.8 m sϪ1,re- uation is more complicated inside a thunderstorm, where spectively. momentum transported by a downdraft could be rep-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2575 resentative of both the environment and a perturbation also approached from the east and rose into the low- produced by the storm. level mesocyclone. However, during tornadogenesis air Davies-Jones and Brooks (1993), while also stressing parcels also approached the mesocyclone from the north the importance of the downdraft in tilting low-level hor- and west (Davies-Jones and Brooks 1993); air parcels izontal vorticity, diagnosed the development of a low- were converging from a broader area during tornado- level rotation that originated as baroclinically generated genesis than earlier. horizontal vorticity. In their numerical thunderstorm Instrumented tower data provide insight into the na- simulation, horizontal vorticity produced at the edge of ture of the vorticity source for the low-level mesocy- the rain-cooled air north of the updraft was tilted into clone; the low-level air was rich in horizontal vorticity. the vertical by the vertical velocity gradient within de- Tower measurements far from the storm indicate a ver- scending air. The vertical vorticity was then ampli®ed tical shear of approximately 1.5 ϫ 10Ϫ2 sϪ1 in the lowest as it entered the region of low-level convergence as- 440 m of the atmosphere (Fig. 18). Assuming that ver- sociated with the updraft. A similar mechanism was tical motions were relatively weak far from the storm, noted in a more idealized numerical experiment by we can infer that the horizontal vorticity (which is equal Trapp and Fiedler (1995). in magnitude, but perpendicular, to the vertical shear) In the Davies-Jones and Brooks (1993) process, the was directed toward the west-southwest. As the storm material curve that eventually encloses the low-level approached, the magnitude of the vertical shear grad- vortex initially has little circulation about it. However, ually increased to 2.8 ϫ 10Ϫ2 sϪ1 at 1634 (Fig. 18) and circulation about the tilted ¯uid circuit builds up owing then peaked at over 3.0 ϫ 10Ϫ2 sϪ1 within the updraft to baroclinicity in the in¯ow (i.e., a temperature gradient (not shown). Possible explanations for the increased pointing in the ϩy direction in Fig. 16). Otherwise, the low-level vertical shear from the far storm to the near subsequent ¯attening and shrinking of the material storm environment include stretching of horizontal vor- curve is similar to that depicted in Fig. 16. ticity in the accelerating storm in¯ow and baroclinic We now return to the Arcadia storm to compare what generation of horizontal vorticity. was observed in nature to the theories of tornadogenesis A gradual temperature drop of 5ЊC(2ЊC) at the sur- deduced from numerical simulations and laboratory ex- face (at 444 m AGL) was experienced by the instru- periments. mented tower during a period of over 1 h between the sunny environment outside the storm and the shaded region with light precipitation immediately upstream of b. Observations of the Arcadia storm the updraft (not shown). Preliminary analyses of su- Analysis of the trajectories of material curves most percells during VORTEX (Rasmussen et al. 1994) have clearly demonstrates the difference between the pretor- shown forward-¯ank storm baroclinicity to be rather nadic and tornadic low-level mesocyclones. Unfortu- diffuse (Rasmussen and Straka 1996; P. Markowski nately, it is dif®cult to obtain quantitative results from 1997, personal communication). The instrumented-tow- a trajectory and circulation analysis applied to real data, er temperature trace of the Arcadia storm also hints that owing to limited spatial resolution and noise in the data. the temperature gradient was spread over a broad region, For the Arcadia storm, no radar scans of the mesocy- but we lack measurements perpendicular to the tower clone are available below 400 m AGL. Data at the low- cross section necessary to con®rm this. Although baro- est level of the Cartesian grid have been extrapolated clinicity may have been weak, air parcels residing in downward from the height of the lowest radar scans; in the broad baroclinic zone for long periods could have reality, the tower measurements con®rm signi®cant ver- still acquired signi®cant horizontal vorticity (P. Mar- tical shear of the horizontal winds in this layer (Fig. kowski 1997, personal communication). 12b). We will focus on features in the ¯ow at 500 m The west-southwest to east-northeast orientation of AGL. However, since many trajectories ascend to this the re¯ectivity contours (Figs. 5, 11) beneath the down- level, the results must be interpreted with caution. We stream anvil of the storm suggests that the temperature anticipate that the trajectory analysis will still provide contours were oriented similarly (i.e., cool air within a sense of parcel origins. The trajectories depicted herein the precipitation region, warm air outside the storm). represent an improvement over those in a preliminary This implies baroclinic generation of horizontal vortic- analysis (Dowell and Bluestein 1996) in that velocity ity directed toward the west-southwest, which is the time derivatives have been incorporated into the current same direction as the preexisting environmental hori- calculations. zontal vorticity and the same direction as the observed In the pretornadic Arcadia storm (Figs. 5b, 17a), all increase in vorticity as the storm approached (Figs. 12b, air parcels entering the mesocyclone at 500 m AGL 18). Thus, low-level mesocyclogenesis may have been originated east of the mesocyclone. The west side of the enhanced by the addition of baroclinic effects to the material curve was tipped upward by the updraft, and preexisting horizontal vorticity (Wicker 1996). thus low-level horizontal vorticity was tilted into the Since the orientation of the tower cross section vertical. through the pretornadic low-level updraft is approx- During the tornadic stage (Fig. 17b), some air parcels imately parallel to the low-level horizontal vorticity

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2576 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 14. Vertical cross sections constructed from a time-to-space conversion of instrumented tower data between 1635 and 1645 CST. Equivalent distance for storm motion of 11.7 m sϪ1 is indicated. (a) Equivalent potential temperature (K) and storm-relative winds. (b) Virtual potential temperature (K) and storm-relative winds. (c) Storm-relative winds and contours of the component of wind normal to the cross section (m sϪ1). Solid (dashed) contours indicate winds into (out of) the page.

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2577

FIG.14.(Continued)

(Figs. 12b, 13), we are able to compute a rough estimate of the rate of vertical vorticity generation by tilting on the in¯ow (east) side of the updraft. Tower measurements at 444 m indicate a vertical velocity gradient of 6 m sϪ1 over a distance of 1 km (Fig. 13). If we assume that half this magnitude of gradient is

FIG. 16. Demonstration of how horizontal vorticity associated with vertical shear can be concentrated into a vortex at low levels. The analysis is carried out in terms of a material circuit in the ¯uid; the process involves tilting of the circuit into the horizontal by a downdraft, followed by convergence of the circuit as it is passed into an FIG. 15. Vertical pro®le of equivalent potential temperature (K) in updraft. The effects of deformation acting on the material curve are the sounding taken from Edmond, Oklahoma, at 1430 CST (cf. Fig. neglected for ease of interpretation. The environmental wind pro®le 2). is indicated at the left.

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2578 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 18. Hodograph for low-level winds measured by the instrumented tower at 1530 (about 50 km ahead of the updraft) and 1634 CST (2 km ahead of the updraft). Each circle represents5msϪ1 of wind speed.

D␨ 3msϪ1 ( ١wϭ(0.025 sϪ1´ϭ␻ Dt hh ΂΃1000 m ϭ 7.5 ϫ 10Ϫ5 sϪ2 . (4) With an updraft-relative ¯ow of about 15 m sϪ1 (Fig. 11b), parcels could pass through the vertical velocity gradient in 70 s, long enough for a vertical vorticity of approximately 0.5 ϫ 10Ϫ2 sϪ1 to be produced by tilting. This vorticity could then be ampli®ed within the updraft by convergence (Brandes 1984b). The tornado formed southeast of the mesocyclone center at 0.5 km AGL in a region of horizontally sheared ¯ow between westerlies on the rear downdraft side of the mesocyclone and storm-in¯ow southeasterlies (Figs. 9 and 10). Storm intercept teams have long observed that the formation of the rear downdraft precedes the formation of signi®cant supercell tornadoes (Lemon and Doswell 1979); the rear downdraft is manifest visually FIG. 17. Material curves traced backward (using storm-relative by a clear slot, which is a relatively bright region within trajectories) from a ring with a 2.5-km diameter centered on the the cloud on the right side (with respect to storm motion) mesocyclone at 0.5 km AGL; x and y as in Fig. 3. Velocities are of the mesocyclone. With the limitations of the radar computed by linear interpolations (in time) between the dual-Doppler volume at the time given in the ®gure and the volume at a previous data, it is dif®cult to diagnose accurately the role the time. Heights (km AGL) of points along the curve are shown. Con- rear downdraft plays in tornadogenesis. However, the tours of vertical vorticity (10Ϫ2 sϪ1) are also shown: (a) followed back proximity of the tornado to the nose where the rear 400 s from 1638 CST, and (b) followed back 400 s from 1704 CST. downdraft penetrates into the mesocyclone (Fig. 9) suggests a dynamical relationship between the two features, although the nature of the relationship is unclear. West- representative of the overall depth of the tower and erly winds that were a part of the horizontally sheared that horizontal vorticity is of magnitude 2.5 ϫ 10Ϫ2 ¯ow in which the tornado was embedded were found sϪ1, then this implies a rate of tilting of horizontal just downstream of the rear downdraft, suggesting that vorticity into the vertical of westerly momentum brought downward in the down-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2579

these ¯ow features developed as the mesocyclone at 2± 3 km was intensifying (Fig. 8) just prior to tornado formation. However, as discussed in section 4, the analyses must be interpreted with caution owing to the many dif®culties in retrieving pressure from dual-Doppler measurements (Hane and Ray 1985). Further diagnosis of rear downdraft formation and its role in tornadogenesis will likely require the use of numerical simulations. FIG. 19. Schematic of features at 1.0 km AGL. Downdrafts (w Յ The traditional explanation for tornado demise is that Ϫ1msϪ1) are shaded lightly, updrafts (w Ն 3msϪ1) are shaded heavily, and mesocyclones (␨ Ն 0.6 ϫ 10Ϫ2 sϪ1) are denoted by solid the out¯ow surges ahead of the primary updraft and curves. The TVS location at 1704 CST is marked. mesocyclone, cutting off the low-level in¯ow into the updraft and, in general, disrupting the balance of mesocyclone processes (Lemon and Doswell 1979; Brooks draft may have played an important role in the devel- et al. 1994). This is consistent with the analysis of the opment of vertical vorticity near the ground, as in the Arcadia storm at least in that the Doppler analysis shows Walko (1993) simulation (Fig. 16). westerly out¯ow winds farther and farther east of the In the Arcadia storm, the rear downdraft appeared to updraft and mesocyclone in successive volumes around separate from the rainy downdraft north of the updraft the time of tornado dissipation (cf. Fig. 5). and with time progress cyclonically around the updraft (Fig. 19). Numerical simulations of supercells indicate 6. Conclusions that the primary storm-scale rear downdraft (north and northwest of the updraft in the Arcadia storm) is driven In many ways, the Arcadia supercell was remarkably by the cooling associated with evaporation of precipi- similar to storms that have been documented in previous tation (Klemp and Rotunno 1983; Brooks et al. 1994). dual-Doppler studies, including the Del City storm As the mesocyclone intensi®es, it wraps precipitation (Brandes 1981). Hopefully the common behavior of the around the updraft (Brooks et al. 1994), which promotes storms indicates that there is indeed a common mech- downdraft development farther behind the updraft (west anism acting to produce many of the tornadoes within of the updraft in the Arcadia storm). isolated supercells. Dynamic effects associated with the intensifying When radar observations began, the Arcadia storm mesocyclone may also lead to downdraft formation. In had a strong updraft and mesocyclone±anticyclone cou- the simulation of Klemp and Rotunno (1983), the ``oc- plet aloft but only a weak mesocyclone at low levels. clusion downdraft'' (that portion of the rear ¯ank down- With time, low-level northerly out¯ow increased on the draft within the low-level mesocyclone and immediately rear side of the storm, a re¯ectivity appendage formed, behind the convergence line of the occluded gust front) and the rear downdraft wrapped cyclonically around the was driven dynamically by the strong circulation. Large updraft. vorticity near the surface with weaker rotation aloft was Tornadogenesis occurred while low-level associated with a downward-directed perturbation pres- mesocyclone vorticity was rapidly increasing. The tor- sure gradient force. nado itself was located southeast of the mesocyclone Pressure retrievals (Hane and Ray 1985) for the tor- center, at the tip of the re¯ectivity hook. During tornadic stage of the Arcadia storm do suggest a downward nadogenesis, convergence was increasing within the perturbation pressure gradient force south and west of low-level mesocyclone. the mesocyclone at low levels (e.g., near y ϭ 45 in Fig. The tornado dissipated while low-level out¯ow 20a), which could be related to the vertical velocity surged well ahead of the primary updraft and meso- minimum that wraps into the southeast side of the mesocyclone. After tornado demise, mesocyclone vorticity cyclone (Fig. 19). This region of downward-directed at low levels exceeded that aloft, and the Arcadia cell pressure forces is apparent only in the retrievals from soon lost supercell characteristics. Another cell devel- 1651 through 1713 (i.e., from tornadogenesis to tornado oped near the leading edge of the out¯ow, but it proved demise). to be short lived. In the numerically simulated storm of Klemp and Before the time of tornadogenesis, the low-level Rotunno (1983), downward acceleration owing to dy- mesocyclone of the Arcadia storm consisted of a rising namic effects was primarily driven by strong vorticity stream of environmental air that entered the storm from z in the the east (Fig. 17a). With time, a rear downdraft formedץ/pץ near the surface. The region of positive Arcadia storm is related both to low pressure near the southwest of the updraft, and air began to converge from surface and to high pressure south and west of the meso- a broader region into the mesocyclone (Figs. 17b, 19). cyclone aloft (at 2±3 km AGL) (Fig. 20). Relatively The development of the rear downdraft in the Arcadia high pressure coincided with dif¯uence in the ¯ow west- storm occurred simultaneously with approximately a dou- southwest of the mesocyclone and con¯uent ¯ow south bling of vorticity in the low-level mesocyclone. The in- and southeast of the mesocyclone (Fig. 20b); both of tensi®cation of the two features may result from a syn-

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2580 MONTHLY WEATHER REVIEW VOLUME 125

FIG. 20. Perturbation pressure (mb) retrieved from dual-Doppler analyses at 1704 and 1710 CST; x and y as in Fig. 3. Contours for negative values are dashed: (a) vertical cross section along x ϭ 12 km [see (b)], and (b) horizontal cross section at 2.5 km AGL.

ergetic interaction. As the rear downdraft intensi®es, it may the updraft. Pressure retrievals for the Arcadia storm sug- aid in strengthening the low-level mesocyclone through gest that the low-level rear downdraft south and southwest downward momentum transport and/or increased conver- of the mesocyclone center during the tornadic stage was gence at the edge of where the downdraft air spreads out in a region with a downward-directed perturbation pressure at the surface. The intensi®cation of the mesocyclone may gradient force associated with relatively high pressure aloft in turn lead to strengthening of the rear downdraft by southwest of the intensifying mesocyclone and low pres- dynamic effects and/or wrapping of precipitation around sure near the surface.

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC OCTOBER 1997 DOWELL AND BLUESTEIN 2581

Perhaps more importantly, a downdraft may be nec- , 1981: Finestructure of the Del City±Edmond tornadic meso- essary to produce the strong, localized, near-surface vor- circulation. Mon. Wea. Rev., 109, 635±647. , 1984a: Relationships between radar-derived thermodynamic ticity characteristic of tornadoes, as hypothesized by variables and tornadogenesis. Mon. Wea. Rev., 112, 1033±1052. Walko (1993) and Davies-Jones and Brooks (1993). The , 1984b: Vertical vorticity generation and mesocyclone suste- Arcadia tornado formed near the nose of the rear down- nance in tornadic thunderstorms: The observational evidence. draft (Fig. 9) in a region of horizontally sheared ¯ow Mon. Wea. Rev., 112, 2253±2269. between southeasterly storm in¯ow and westerly out- , R. P. Davies-Jones, and B. C. Johnson, 1988: Streamwise vorticity effects on supercell morphology and persistence. J. Atmos. ¯ow downstream of the rear downdraft (Fig. 10). The Sci., 45, 947±963. evidence, although circumstantial, suggests an active Brewster, K. A., 1984: Kinetic energy evolution in a developing se- role of the rear downdraft in tornadogenesis. vere thunderstorm. M.S. thesis, School of Meteorology, Uni- The most novel aspect of this study is that it is the versity of Oklahoma, 171 pp. [Available from School of Me- ®rst to synthesize independent measurements of the 3D teorology, 100 East Boyd, Norman, Oklahoma, 73019.] Brooks, H. E., C. A. Doswell, and R. B. Wilhelmson, 1994: The role low-level ¯ow in and near the mesocyclone of a su- of midtropospheric winds in the evolution and maintenance of percell. Doppler analyses provide an overall perspective low-level mesocyclones. Mon. Wea. Rev., 122, 126±136. of the wind ®eld within the Arcadia storm and instru- Brown, R. A., L. R. Lemon, and D. W. Burgess, 1978: Tornado mented tower measurements resolve some of the ®ner- detection by pulsed Doppler radar. Mon. Wea. Rev., 106, 29± scale details of the low-level ¯ow and thermodynamic 38. Browning, K. A., 1964: Air¯ow and precipitation trajectories within structure. severe local storms which travel to the right of the winds. J. Tower data show that the low-level environment of Atmos. Sci., 21, 634±639. the Arcadia storm was rich in horizontal vorticity; this Burgess, D. W., 1982: Lifecycle of the Wichita Falls tornadic storm. vorticity entered the pretornadic storm from the east in Preprints, 12th Conf. on Severe Local Storms, San Antonio, TX, an approximately streamwise fashion and was tilted up- Amer. Meteor. Soc., 441±443. Carter, J. K., 1970: The meteorologically instrumented WKY-TV ward into the low-level mesocyclone. Instrumented tow- tower facility. NOAA Tech. Memo. ERLTM-NSSL 50, 18 pp. er measurements indicate low-level vertical shear im- [NTIS COM-71-00108.] mediately upstream of the updraft almost twice as great Cressman, G. P., 1959: An operational objective analysis system. as observed farther away in the environment. This in- Mon. Wea. Rev., 87, 367±374. crease may have resulted from enhancement of the low- Davies, J. M., and R. H. Johns, 1993: Some wind and instability level horizontal vorticity by baroclinic generation and/or parameters associated with strong and violent tornadoes. 1. Wind shear and helicity. The Tornado: Its Structure, Dynamics, Pre- stretching of horizontal vorticity in the storm in¯ow. diction and Hazards, Geophys. Monogr., No. 79, Amer. Geo- phys. Union, 573±582. Acknowledgments. The Doppler radar data were an- Davies-Jones, R. P., and H. E. Brooks, 1993: Mesocyclogenesis from alyzed using the following software developed at the a theoretical perspective. The Tornado: Its Structure, Dynamics, National Center for Atmospheric Research: RDSS (data Prediction and Hazards, Geophys. Monogr., No. 79, Amer. Geo- phys. Union, 105±114. REORDER CEDRIC editing), (objective analysis), and , D. W. Burgess, and M. Foster, 1990: Test of helicity as a forecast (wind synthesis). NCAR is supported by the National parameter. Preprints, 16th Conf. on Severe Local Storms, Kan- Science Foundation. Graphics were generated using anaskis Park, AB, Canada, Amer. Meteor. Soc., 588±592. ZXPLOT, developed at the University of Oklahoma by Doswell, C. A., III, and D. W. Burgess, 1993: Tornadoes and tornadic Ming Xue. storms: A review of conceptual models. The Tornado: Its Struc- We are indebted to Don Burgess, who suggested that ture, Dynamics, Prediction and Hazards, Geophys. Monogr., No. 79, Amer. Geophys. Union, 161±172. we take a look at this case. Jerry Wardius provided a Dowell, D. C., and H. B. Bluestein, 1996: Dual-Doppler analysis of copy of the radar and instrumented tower data. David a tornadic supercell: The Arcadia, OK storm of 17 May 1981. Stensrud made upper-air and surface data available to Preprints, 18th Conf. on Severe Local Storms, San Francisco, us. Carl Hane assisted with the pressure retrieval and CA, Amer. Meteor. Soc., 413±417. Jeff Trapp provided helpful comments on some of the , , and D. P. Jorgensen, 1997: Airborne Doppler radar analysis of supercells during COPS-91. Mon. Wea. Rev., 125, 365± topics in this paper. 383. This work was funded by NSF Grant ATM-9302379. Doviak, R. J., P. S. Ray, R. G. Strauch, and L. J. Miller, 1976: Error estimation in wind ®elds derived from dual-Doppler radar measurement. J. Appl. Meteor., 15, 868±878. REFERENCES Fujita, T. T., G. S. Forbes, and T. A. Umenhofer, 1976: Close-up view Barnes, S. L., 1970: Some aspects of a severe, right-moving thun- of 20 March 1976 tornadoes: Sinking cloud tops to suction vor- derstorm deduced from mesonetwork rawinsonde observations. tices. Weatherwise, 29, 116±131. J. Atmos. Sci., 27, 634±648. Goff, R. C., 1976: Vertical structure of thunderstorm out¯ows. Mon. , 1973: Mesoscale objective map analysis using weighted time- Wea. Rev., 104, 1429±1440. series observations. NOAA Tech. Memo. ERL NSSL-62, 60 pp. Grasso, L. D., and W. R. Cotton, 1995: Numerical simulation of a [NTIS COM-73-10781.] tornado vortex. J. Atmos. Sci., 52, 1192±1203. Brady, R. H., and E. J. Szoke, 1989: A case study of nonmesocyclone Guerrero, H., and W. Read, 1993: Operational use of the WSR-88D tornado development in northeast Colorado: Similarities to wa- during the November 21, 1992 southeast Texas tornado outbreak. terspout formation. Mon. Wea. Rev., 117, 843±856. Preprints, 17th Conf. on Severe Local Storms, St. Louis, MO, Brandes, E. A., 1978: Mesocyclone evolution and tornadogenesis: Amer. Meteor. Soc., 399±402. Some observations. Mon. Wea. Rev., 106, 995±1011. Hane, C. E., and P.S. Ray, 1985: Pressure and buoyancy ®elds derived

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC 2582 MONTHLY WEATHER REVIEW VOLUME 125

from Doppler radar in a tornadic thunderstorm. J. Atmos. Sci., K. K. Wagner, R. B. Wilhelmson, and J. B. Klemp, 1981: The 42, 18±35. morphology of several tornadic storms on 20 May 1977. J. At- Johnson, K. W., P. S. Ray, B. C. Johnson, and R. P. Davies-Jones, mos. Sci., 38, 1643±1663. 1987: Observations related to the rotational dynamics of the 20 Rotunno, R., and J. Klemp, 1985: On the rotation and propagation May 1977 tornadic storms. Mon. Wea. Rev., 115, 2463±2478. of simulated supercell thunderstorms. J. Atmos. Sci., 42, 271± Klemp, J. B., and R. B. Wilhelmson, 1978: Simulations of right- and 292. left-moving storms produced through storm splitting. J. Atmos. Taylor, W. L., Ed., 1982: 1981 spring program summary. NOAA Tech. Sci., 35, 1097±1110. Memo. ERL NSSL-93, 97 pp. [NTIS PB82-244757.] , and R. Rotunno, 1983: A study of the tornadic region within Trapp, R. J., and B. H. Fiedler, 1995: Tornado-like vortexgenesis in a supercell thunderstorm. J. Atmos. Sci., 40, 359±377. a simpli®ed numerical model. J. Atmos. Sci., 52, 3757±3778. Leise, J. A., 1981: A multidimensional scale-telescoped ®lter and , and E. D. Mitchell, 1995: Characteristics of tornadic vortex data extension package. NOAA Tech. Memo. ERL WPL-82, 18 signatures detected by WSR-88D radars. Preprints, 27th Conf. pp. [NTIS PB82-164104.] on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 211±212. Lemon, L. R., and C. A. Doswell, 1979: Severe thunderstorm evo- Wakimoto, R. M., and J. W. Wilson, 1989: Non-supercell tornadoes. lution and mesocyclone structure as related to tornadogenesis. Mon. Wea. Rev., 117, 1113±1140. Mon. Wea. Rev., 107, 1184±1197. , W. C. Lee, H. B. Bluestein, C. H. Liu, and P. H. Hildebrand, Lewellen, W. S., 1993: Tornado vortex theory. The Tornado: Its Struc- 1996: ELDORA observations during VORTEX 95. Bull. Amer. ture, Dynamics, Prediction and Hazards, Geophys. Monogr., No. Meteor. Soc., 77, 1465±1481. 79, Amer. Geophys. Union, 19±39. Walko, R. L., 1993: Tornado spin-up beneath a convective cell: Re- Magsig, M. A., and D. W. Burgess, 1996: A vorticity and divergence quired basic structure of the near-®eld boundary layer winds. analysis relating to tornadogenesis as seen by a WSR-88D radar. The Tornado: Its Structure, Dynamics, Prediction and Hazards, Preprints, 18th Conf. on Severe Local Storms, San Francisco, Geophys. Monogr., No. 79, Amer. Geophys. Union, 89±95. CA, Amer. Meteor. Soc., 418±422. Ward, N. B., 1972: The exploration of certain features of tornado O'Brien, J., 1970: Alternative solutions to the classical vertical ve- dynamics using a laboratory model. J. Atmos. Sci., 29, 1194± locity problem. J. Appl. Meteor., 9, 197±203. 1204. Rasmussen, E. N., and J. M. Straka, 1996: Mobile mesonet obser- Weisman, M. L., and J. B. Klemp, 1984: The structure and classi- vations of tornadoes during VORTEX-95. Preprints, 18th Conf. ®cation of numerically simulated convective storms in direc- on Severe Local Storms, San Francisco, CA, Amer. Meteor. Soc., tionally varying wind shears. Mon. Wea. Rev., 112, 2479±2498. 1±5. Wicker, L. J., 1996: The role of near surface wind shear on low-level , , R. Davies-Jones, C. A. Doswell III, F. H. Carr, M. D. mesocyclone generation and tornadoes. Preprints, 18th Conf. on Eilts, and D. R. MacGorman, 1994: Veri®cation of the origins Severe Local Storms, San Francisco, CA, Amer. Meteor. Soc., of rotation in tornadoes experiment: VORTEX. Bull. Amer. Me- 115±119. teor. Soc., 75, 995±1006. , and R. B. Wilhelmson, 1995: Simulation and analysis of tornado Ray, P. S., C. L. Ziegler, W. Bumgarner, and R. J. Sera®n, 1980: development and decay within a three-dimensional supercell Single- and multiple-Doppler radar observations of tornadic thunderstorm. J. Atmos. Sci., 52, 2675±2703. storms. Mon. Wea. Rev., 108, 1607±1625. Wilson, J. W., 1986: Tornadogenesis by nonprecipitation-induced , B. C. Johnson, K. W. Johnson, J. S. Bradberry, J. J. Stephens, wind shear lines. Mon. Wea. Rev., 114, 270±284.

Unauthenticated | Downloaded 09/25/21 04:09 PM UTC