THUNDERSTORM-SCALE VARIATIONS of ECHOES ASSOCIATED with LEFT-TURN TORNADO FAMILIES Gregory S. Forbes the University of Chicago 1

Home , Hook echo

Reprh\ud from Preprint VolUM: : Ten th Confereac.e o n Severe t.oce l Stor... OC: t . 18·21, 1 977 ; OMhl, >ft. Publtehed by American Heteorologt cet Soctety. lo•ton, H.uH chuntu. THUNDERSTORM-SCALE VARIATIONS OF ECHOES ASSOCIATED WITH LEFT-TURN TORNADO FAMILIES

Gregory s. Forbes The University of Chicago

1. INTRODUCTION Browning and Donaldson (1963), and Fujita One of the largest tornado outbreaks (1972), among others. in history occurred on 3 April 1974. Fol This paper is a continuation of lowing the outbreak, an intensive damage Forbes (1975). The emphasis in that paper survey (Fujita, 1975a) confirmed the exist was to establish a s tatistical estimate of ence of 148 tornadoes, whose paths are the validity of using echo s hape as a tor shown in Figure 1. The outbreak focused nado indicator. An elaboration on this attention on the periodic production of work is presented in Section 2. Sections tornadoes by a long-lived thunderstorm, 3 and 4 represent the main emphasis of this creating a tornado family. Of particular paper: · associating changes in echo features interest were several spectacular left with tornado touchdowns, liftoffs, and turns. turn tornado families and the preponderance of left-turn tornadoes . 2. DISTINCTIVE ECHO STATISTICS As noted by Darkow and Roos (1970) and others, the periodic formation of tor Two echo configurations that are nadoes in a tornado family tantalizingly known to be associated with tornadoes are suggests a concurrent cyclic variation of the hook echo and the line-echo-wave some thunderstorm feature . This paper patte rn (LEWP). Nolen (1959) defined a shows that such a thunderstorm cycle exists LEWP as "a configuration of radar echoes in in conjunction wi th left -turn tornado fam which a line of echoes has been subj ected ilies. This cycle is revealed in the r adar to an acceleration along one portion and/ or echo by a change of· orientation of t he hook a deceleration along that portion of the and successive formations of new hooks. line immediately adjacent, with a resulting Other studies have explored poss ible cyclic thunderstorm features i n r elation to tornadoes. Foster (1973) studied a tornado family associated with an echo protuberance (rather than a classic hook) . He found that the echo grew in bursts with a 46-minute periodicity. Following such a burst, the echo orientation shifted cyclon ically and the echo made a right turn in movement. One tornado formed just after the cyclonic shift. The other touchdowns were less precisely known. Doppler radar observations by Donaldson (1971) and Burgess (1976), for example, indicate that a tornado forms after a mesocyclone signature has descended to the surface. There has been considerable research on the behavior of the thunders t orm top as a possible indicator of tornado pro duction. In general, these studies have concluded that tornadoes occur as the cloud top descends or i s at a minimum height. These studies include Donaldson (1958), Fig. 1. Tracks of the 148 tornadoes of April 3, 1974. From Fu jita (1975a ).

497 sinusoidal mesoscale wave pattern in the by one radar to another. These were ini line". Hamilton (1970) found that the LEWP tially counted as new echoes if there was was associated with a mesolow and that the a detectability gap during the crossover; acceleration of the south part of the line but now each echo is counted only once. was accomplished by a spreading out of the The tracks of the distinctive echoes, mesohigh. Cook (1961) observed that the from echo origin to termination, are shown severe weather occurred with the strongest in Figure 3. cell, in the vicinity of the mesolow. The hook echo has been recognized as a potential tornado indicator since first reported by Stout and Huff (1953). Quanti tative evaluations of the reliability of the hook echo as a tornado indicator. have been performed only in limited studies, however. Freund (1966) found that 6 of 13 tornadic storms near NSSL in 1964 had hook echoes. Sadowski (1969) studied documented hook echoes from 1953 to 1966. Data tabu lated in his appendices indicate that 40 of the 46 hook echoes were tornadic. Using Doppler radar, Burgess (1976) reports that from 1971 to 1975, 23 of 37 meso cyclones detected near NSSL were tornadic. Three types of echoes which commonly produced tornadoes were classified as "distinctive" in this study, on the basis of echo shape. These distinctive edhoes, shown in Figure 2, are hook echoes, LEWP's, and echoes with appendages on the southwest of the echo and oriented at an angle of at least 40 degrees to the south of the echo movement. The study area was described by Forbes (1975). Briefly, it was composed of all WSR-57 radars operating in the 125 nautical mile range mode. There were 55 distinctive echoes in the study. This represents a decrease of PATHS OF five from Forbes (1975), due to reanalysis DISTINCTIVE ECHOES of the echoes which crossed from coverage

Fig. 3. Paths of distinctive echoes.

Table l reveals that 65%· of the distinctive echoes produced tornadoes. Also, 65% of the echoes which produced tornadoes were of distinctive shape. (The identical percentages are coincidental.) Distinctive echoes produced 27 of the 28 tornado families in the study. About half of the distinctive echoes produced a tor nado family.

DISTINCTIVE ECHOES 55 Tornadic 36 (65%) Non-tornadic 19 (35%)

TORNADIC ECHOES 55 Fig. 2. Types of distinctive echoes. In the Distinctive 36 (65%) diagram, A refers to appendage and Non-distinctive 19 (35%} HL refers to hooklike--a very pronounced appendage but not of a Table 1. Distribution of echoes classic hook shape. and tornadoes.

498 Tornadoes produced from distinctive other extreme, one might hypothetically echoes tended to be stronger and lasted issue warnings of one-minute duration at longer than those from non-distinctive one-minute intervals, based upon whether the echoes. Table 2 shows that 85 of the total echo was distinctive at that moment. This 105 tornadoes (81%) in the study were is defined here as an instantaneous warning. produced from distinctive echoes, including The "real" verification scheme lies between all F4 and FS tornadoes. The mean F-scale these two schemes, and is determined by the of tornadoes from distinctive echoes was public's concept of how close a tornado 3.26 while those from non-distinctive must strike in order to justify a warning echoes had mean of 1.88. 35% of the dis at their location. tinctive echoes produced an F4 or FS tor nado. The median duration of tornadoes Xndefillit• wami!!g from distinctive and non-distinctive echoes p .o . D. • lhmber of tomadot• produced frsp 4i•$ip£tive ecbO!• was 20 and 5 minutes, respectively. Tot.al nWllber of torn.-c1o9•

p .A.a. • !UP!r of non-tornadic but 41et1qctive ecl\O!f TORNADOES FROM TORNADOES FROM NON Tot:al n'81ber of dietinc:tiv• ed\oea · DISTINCTIVE ECHOES DISTINCTIVE ECHOES PO S ( 63 ) FO 6 (303 ) P.O.D. • •-r of tornadoeo p

Donaldson et al. (1975a,b) to include these i!: F5 factors. i!:F4 The definitions of Table 3 represent two extremes. If one is not concerned with i!:F3 false alarms, when the echo first becomes i!:F2 distinctive he might hypothetically issue a warning over an area sufficient to cover the remaining lifetime of the echo. This is defined here as an indefinite warning, and verifies if a tornado occurs for any Fig. 4. POD and FAR for indefinite and portion of the warning duration. At the instantaneous warnings.

499 Tpe generality of the FAR and POD occurring to the rear may belong to a for the April 3 study is questionable. separate category of thunderstorm-tornado Both the data above (6~%) and that of interaction. Burgess (19761 62%) indicate that 60 to 65% of detectable mesocyclones produce In Hook Rear of Hook tornadoes. Thus the above false alarm LEWP rates (35% for indefinite warnings and IFormation 46 7 51% for one-hour warnings) seem reasonable. Liftoff 42 On the other hand, more research is .needed 7 to determine what fraction of all tornadoes Table 4. Location of tornadoes relative are produced from distinctive echoes. to echoes at short ranges. Operational detectability is also a problem, wit~ distinctive echoes going undetected at long ranges. In the April Table 5 gives a more detailed break 3 study, the overall detectability.was down of the location of touchdowns and liftoffs. The locations given are with only about 74%, due to gaps in coverage respect to the hook for the categories of (where the nearest radar was over 100 n.mi./ 185 km .away). hook and rear, and with respect to the echo center for LEWP. 76% of the tor nadoes forming in the hook were. in the 3. ECHO CHANGES AND TORNADO OCCURRENCES SSE to SSW quadrant of the hook, often The appendages and hooks of the within an asc or spiral near the end of distinctive echoes were not steady in shape the hook. The tornadoes forming in the or orientation. Only 45% of the distinctive hook tended to move approximately with echoes ever assumed a classic hook shape. the hook, except near liftoff, although Non-tornadic echoes possessed a hook even the hook often changed orientation. less frequently. In total, the echo assumed Near liftoff, left-turn tornadoes. also a hook shape for only 20% of the time that changed position within the hook, moving it was distinctive. At times, the echo westward or northwestward. momentarily lost its distinctive shape. Table 5 also presents a summary of Changes in appendage and hook orientation the positions of the 14 LEWP tornadoes. were usually accompanied by tornado turns, Many occurred near the circulation center, as discussed in Section 4. and 85% occurred either near the main When studied a posteriori, distinctive echo center or southwest to west of it. echoes which produced tornadoes were distinctive longer than those which were Location of Formation non-tornadic: 145 versus 62 minutes in Book Rear LEWP median duration. Tornadic distinctive E echoes also had longer lives: 330 versus ESE 2% 217 minutes in median. life. There was a SE 9 general tendency for distinctive echoes SSE 17 with lives of 300 to 420 minutes to pro s 46 duce the strongest (F5) tornadoes. No SSW 13 7% SW 13 43% 14 echo with life of less than 300 minutes WSW 14 14 produced an F4 or F5 tornado. on an w 36 individual basis, however, it was difficult WNW 14 to distinguish a tornadic from a non NW 29 7 Center 21 tornadic distinctive echo simply by Unknown studying its shape. Location of Liftoff After calibrating the echo positions Book Rear through use of known ground clutter, the E tornado was located with respect to the ESB echo. (Figure 7 illustrates the locations SE 7% of some of the ground clutter used in cali SSE 10 s 38 brating the Cincinnati radar.) . Table 4 sum SSW 2 marizes these observations. The locations SW 19 are subject to some error, introduced due WSW 7 to some uncertainty in the tornado times, w 17 29% possibly resulting in an overestimate of the WNW NW 7l percentage of tornadoes occurring within · the hook. Of the 53 touchdowns from hook Table 5. Formation and liftoff of tor- echoes at short. ranges, 87% occurred within nadoes relative to the echo, the hook, rather than t9 its rear. The 13% by categories. 500 There ·were often specific echo events making left turns. All of the strong left coincident in time with tornado touchdowns turns and 80% of all the left turns were and liftoffs. 46 touchdowns and liftoffs, accompanied by that event. No-turn and whose times were well documented, were used right-turn tornadoes were usually accom in compiling Tables 6 and 7. panied by echo mergers or weakening ap Tornado touchdowns were primarily pendages. associated with intensification of existing Strong appendages ( 3 5% of touchdowns) or me.rgers Left Ro Left lli9ht 'fllrn on existing appendages (17%). Presumably Turn Turn Turn these cases represent increases in vor Blon9ation anti/or ticity. About 30% of the touchdowns were bump on echo r• ar 10°" 43" 12" '" not accompanied by a noticeable echo event. AP1>9nda9e -alc• n.o 0 7 12 15 See Table 6 for other echo events. Merqer 7• 7 18 31 LlllP 0 29 47 31 There was a definite pattern associ Ro event detected 0 14 12 15 ated with the liftoff of left-tum tor Total tornadoea 14 nadoes: the orientation of the appendage 14 17 13 shifted toward the southwest or west · (48% of the liftoffs). Sometimes a small TOltllADO Bltl!AalOWll BY SUB-Gk>OPS bump moved up the rear of the appendage lllo119"tion and/or bump Appendaqe weakens or the echo (15% of the liftoffs). See Stron9 left 14 (61") Stronq left 0 ( °"' Left 6 (26") Left 1 (2°") Table 7 for other echo events. Ro turn 2 ( 9") Ro tum 2 (40l') !light 1 ( 429 Right ~ Total 23 Total s BCBO EVENT NUMBER PERCENT Merger Line Bebo Wave Pattern Stron9 left 1 (ll") Stron9 left O ( °"' Appendage intensificatio.n Left 1 (11") Left 4 (25") or formation 16 35% Bo turn 3 (33") Ro turn 8 ( 50'9 Right · ~ Jtight ...!...ilm rntensification southwest, Total 9 Total 16 3 7 but no appendage present Bo Bvent Stron9 left 0 ( 0") Merger of echoes 2 4 Left 2 (33") No turn 2 (33") * One liftoff involved Merger on appendage 8 17 Right -1...ilm both evente. Total 6 Merged echo weakens l 2 Interaction without merger 3 7 Table 8. Tornado turns and echo events. Not noticeable ....!L 30 The Hamburg echo underwent probably Total 46 Touchdowns the most spectacular orientation change, Table 6. Echo events and tornado touch as the Hamburg tornado was occurring. downs. Figure 5 shows that the orientation of the hook changed from about 200 degrees to ECHO EVENT NUMB BR PERCENT about 245 degrees in a 30-minute period. This orientation change was accompanied by Appendage elongation to 22 48% an increase in the length of the neck por the southwest or west tion of the hook, as graphed in Figure 6. Bump moves up rear of 7 15 The orientation gradually changed, and · .the echo then accelerated in bursts after 1515 CDT, Appendage weakens without with the bursts occurring during the left 5 11 elongation turn portion of the tornado. Figure 7 Merger 9 20 relates the location of the tornado along its path with the tornado position in the rnteracting echo weakens 4 9 echo, shown in Fig. 5. Not noticeable _6_ 13 The left turn of the Hamburg tor Total 53 Liftoffs nado was only partially caused by the Table 7. Echo events and tornado liftoffs. orientation change of the hook. It can be seen in Fig. 5 that the tornado also changed position within the hook. Until 1535, the tornado was located in the oval-shaped echo forming the southeast portion of the 4. ECHO CHANGES AND TORNADO TURNS hook. After 1539, the tornado moved northwestward toward the neck of the hook. A breakdown of the echo events This movement is diagrammed in Figure 8. associated with tornado ~urns is shown in Similar movement of the tornado relative to Table 8. An orientation change, which is the hook has been cited by Fujita (1975b) referred to as an elongation in the table, by the Xenia, Brandenburg, and Sayler was primarily associated with tornadoes Park tornadoes of April 3, 1974. 501 N +

+ + \ + GROUND + CLUTTER REFERENCE 1520 1529

10 20 30 4 0 50 KM I I I I I I I I I UJl ISU 10 20 30 N. Ml. Fig. 7. Tracks of the Hamburg tornado and the Depauw family.

'LIFTOFF

·~ 43 41 1515 uu

____ JI ____ J2 N. MI. I I Fig. 5. The Hamburg 38 echo, from CVG radar. Fig. 8. Movement of the Hamburg tornado relative to the hook.

U4J The Depauw tornado f amily was a family of six left-turn tornadoes. Figure 9 illustrates the thunderstorm cycle asso ciated with the left-turn tornado family. The locations of. the tornadoes along their paths are shown in Fig. 7. At 1549, the Madison tornado was within the hook echo~ At 1554 the hook appeared to wrap up, and it began to shift northwestward, as evident by 1559. After 1602, the appendage no longer had a hook shape. The old appendage continued to move northwestward until about 1614, when it began to dissipate. The Madison tornado moved basically with the appendage, and slightly northwestward rel ative to it. Meanwhile, the Bear Branch tornado touched down in an ehco-free region south of the Madison appendage at 1604. A hook began forming around the Bear Branch tor nado at 1609 and was distinct after 1614. It appears that this left-turn tornado Fig. 6. Orientation and neck length of family was associated with an unsteady the Hamburg echo. thunderstorm. Each tornado was associated with its own hook, which changed orienta- 502 tion and was succeeded by a new hook to by eyewitnesses that there is often clear the south. The Xenia and Acworth echoes, sky immediately west of the tornado. This among others, behaved similarly. There may represent a different category of was no evidence that the tornado families tornado ,origin from the unsteady-hook mode were produced -by several tornado cyclones described above in conjunction with the revolving about a conunon center, as pro Depauw family. posed by Agee et al. (1976).

UIO 151' uo UH

I~. PARKER

14, .p"""f 1410 "--....__ "-., l::_NNA~D I '"'" M FOUN TAINTOWN O Co,_ M- CVG + I I I I 50 N Ml 1549 1604 Fig. 10. The Parker-family echo, from CVG radar. Fig. 9. The Depauw family echo, from CVG radar. 5. DISCUSSION Based upon the s~udy of echo shape and its changes, and tornado location relative to the echo, it appears that there may be at least three classes of tornado 161' origin associated with hook echoes: l. A "steady" hook mode, .producing By contrast with the Depauw echo, no-turn or perhaps right-turn the Parker echo was quite dis simi lar in tornadoes, where the hook does terms of tornado locations relative to the not change shape or orientation: echo. This difference might be related 2. An "unsteady" hook mode, pro to the difference in orientati ons of the ducing left-turn tornado families, families, which can be seen in Figure l. where the hook changes orienta The Parker family was oriented along a tion and a new hook forms in more northerly track. association with each successive Figure 10 shows that the Parker tornado: family tornadoes occurred primarily on the 3. A "rear-subcell" mode, producing rear of the echo. In par ticular, the left-turn tornado families, where Kennard tornado at 1518 and the Parker tornadoes move along the rear of tornado at 1549 appeared to be associated a hook which does not undergo an with protube rances on the rear of the echo. orientation change. Such an occurrence was a l so suspected by The unsteady-hook mode confirms Garret t and Rockney (1962) . The locations. suspicions that periodic tornado produc of the Parker-fami ly tornadoe s along their tion is related to a thunderstorm-scale paths is also shown in Fig. 10. cycle. The mechanism of this cycle is not It became apparent in relating radar understood. Closing of the hook seems to echoes and tornado times that some other play a key role in the process, however. tornadoes were not occurring inside a As the hook becomes overdeveloped, the classic hook, but rather on the rear edge hook wraps up into a spiral, closing off of the echo. (This type of occurrence has a segment of the WER. When this occurs, been tabulated in Tables 4 and 5.) This the closed-off spiral appears to drift observation is consistent with observations northwestward. Garrett and Rockney (1962) also observed this phenomenon. Presumably. 503 the closing of the hook disrupts the Donaldson, R.J., 1971: Doppler radar iden mesocyclone and the inflow/updraft of the tification of damaging convective storms thunderstorm. This mode may be associated by plan shear indicator. Preprints, with a vigorous flanking cell. This type Seventh Conf. Severe Local Storms, of thunderstorm may be of the severely AMS, 71-74. sheared or multicell type, defined by Donaldson, R.J., R.M. Dyer, and M.J. Kraus, Marwitz. (1971). 1975a: operational benefits of meteor The rear-subcell mode probably rep ological Doppler radar. AFCRL-TR-75- resents tornado formation in a cell of a 0103, Air Force Cambridge Research Lab., multicell or severely-sheared thunderstorm. 26pp. The tornado is probably associated with the updraft of an individual cell, mov-ing Donaldson, R.J., R.M. Dyer, and M.J. Kraus, along the rear of the thunderstorm, and l975b: An objective evaluator of moving to the left of the overall storm techniques for predicting severe weather propagation. events. Preprints, Ninth Conf. Severe Local Storms, AMS, 321-326. Forbes, G.S., 1975: Relationship between tornadoes and hook echoes on April 3, 1974. Preprints, Ninth Conf. Severe Local Storms, AMS, 280-285. Acknowledgement:- Research sponsored by NASA under Grant NGR 14-001-008, NOAA under Grant 04-4-158-1 and by NRC under Foster, H., 1973: Tornado echo study Contract No. AT(49-24)-0239o The author with VIP. Preprints, Eighth Conf. is grateful to Professor T. T. Fujita Severe Local Storms, AMS, 161-164. for his support. Freund, R. F., 1966: Radar echo signature of tornadoes. Proceedings, 12th Conf. Radar Meteor., AMS, 362-365. Fujita, T.T., 1972: Tornado occurrences related to overshooting cloud-top heights as determined from ATS pictures. SMRP Res. Paper. No. 97, ·univ. of Chicago, 32pp. Fujita, T.T., 1975a: Superoutbreak tor REFERENCES nadoes of April 3-4, 1974. Final edition map, The Univ. of Chicago. Agee, E.M., J.T. Snow, and P.R. Clare, Fujita, T.T., 1975b: New evidence from 1976: Multiple vortex features in the April 3-4, 1974 tornadoes . Preprints, tornado cyclone and the occurrence of Ninth Conf. Severe Local Storms, AMS, tornado families. Mon. Wea. Rev., 248-255. 104, 552-563. Garrett, R.A. and v.D. Rockney, 1962: Browning, K.A. and R.J. Donaldson, 1963: Tornadoes in northeastern Kansas, Airflow and structure of a tornadic May 19, 1960. Mon. Wea. Rev., 90, storm. J.Atmos.Sci., 20, 533-545. 231-240. Burgess, D.W., 1976: Single Doppler vortex Hamilton, R.E., 1970: Use of detailed recognition: . part I - mesocyclone intensity radar data in mesoscale signatures. Preprints, 17th Conf. surface analysis .of the July 4, 1969 Radar Meteor., AMS, 97-103. storm in Ohio. Preprints, 14th Conf. Cook, B.J., 1961: Some radar LEWP obser Radar Meteor., AMS, 339-342. vations and associated severe weather. Marwitz, J.D. , 1971: Severe storm types Proceedings, 9th Weather Radar Conf., in relation to. wind shear. Preprints, AMS, 181-185. Seventh Conf. Severe Local Storms, Darkow, G.L. and J.C. Roos, 1970: Multiple AMS, 240-243. tornado-producing thunderstorms and Nolen, R.H., 1959: A radar pattern asso their apparent cyclic variations in ciated with tornadoes. Bull. Amer. intensity. Preprints, 14th Conf. Radar Meteor. Soc., 40, 277-279. Meteor., AMS, 305-308. Sadowski, A., 1969: Size of tornado Donaldson, R.J., 1958: Analysis of severe warning area when issued on basis of convective storms observed by radar. radar hook echo. ESSA Tech. Mem. J.Meteor., 15, 44-50. WBTM FCST- 10, 26pp.

504