2052 MONTHLY WEATHER REVIEW VOLUME 122Evolution and Morphology of Two Splitting Thunderstorms with Dominant Left-Moving Members RODGER A. BROWNNOAA, Environmental Research Laboratories, National Severe Storms Laboratory, Norman, Oklahoma REBECCA J. MEITLNCooperative Institute for Research in Environmental Sciences, Boulder, Colorado(Manuscript received 14 June 1993. in final form 28 January 1994)ABSTRACT During the late afternoon and early evening of 27 June 1989, three splitting thunderstorms formed overStanding Rock Indian Reservation in the southern portion of the North Dakota Thunderstorm Project area. Thefirst two storms are the subject of this study. The entire life cycles of both storms were documented using asingle ground-based Doppler radar. Radar reflectivity signatures of updraft summits and Doppler velocity signatures of divergence near storm top were used to deduce updraft evolution within the storms. Dual-Dopplerradar observations from a ground-based radar and an airborne Doppler radar provided fragmentary documentation of the storms' life cycles. The splitting storms on that day were unusual in two distinct ways: (a) the left members of the splitting stormswere the dominant and longer-lasting ones, and (b) none of the deduced updrafts were collocated with centersof vorticity signatures that would have indicated updraft rotation. Both of the left-moving storms had 10 sequential primary updrafts, whereas their right-hand counterparts had 3 or 4 primary updrafts. Initial formationof the right-flank updrafts lagged behind the initial formation of the !eft-flank updrafts by 40-70 min. All theindividual updraft summits moved in the general direction of the mean wind. Sequential updraft developmenton the left and right flanks of the storms suggested that expanding gust fronts provided the propagationalcomponent of storm motion.1. Introduction In his radar study of thunderstorms in the vicinity ofMontreal, Hitsehfeld (1960) was among the first tonote in the radar reflectivity pattern of an isolated thunderstorm that the storm can split into two storms moving along divergent tracks. Although many studies ofstorm splitting have been undertaken since then, onlyin a few of them have Doppler radar measurementsbeen made within one or both members of the splittingstorm pair. The Doppler radar data typically reveal acyclonically rotating updraft in the dominant rightmoving storm and an anticyclonically rotating updraftin the weaker left-moving storm. These and other characteristics of splitting storms are discussed in section 2. This study documents the entire life cycles of twosplitting thunderstorms based on ground-based singleDoppler radar coverage. Limited portions of the life Corresponding author address: Dr. Rodger A. Brown, NOAA,National Severe Storms Laboratory, 1313 Halley Circle, Norman,OK 73069.cycles were based on dual-Doppler radar coveragefrom the ground-based radar and an airborne Dopplerradar. The storms formed over Standing Rock IndianReservation along the border between North and SouthDakota on 27 June 1989 during the North DakotaThunderstorm Project (Boe et al. 1992). The variousdata sources and analysis techniques used in this studyare discussed in section 3. The environment withinwhich the storms formed is discussed in section 4. The two splitting storms (as well as a third one thatis the subject of another study) that occurred on 27 June1989 were unusual for two reasons. Foremost, the leftmember of each split was the taller and longer-lastingmember, in contrast to the typical situation where thefight member is the dominant one. Second, none of theupdrafts in the two storms studied here appeared torotate (that is, updraft collocated with vertical vorticity) during their mature stage. However, a pair of counterrotating vortices (rotating about vertical axes) developed on the midaltitude lateral flanks of most updrafts. The finescale temporal resolution (Doppler radarvolume scans every 3-5 min) in this study permitteddocumentation of discrete propagation of updraft sumc 1994 American Meteorological SocietySEPTEMBER 1994 BROWN AND MEITiN 2053mits within the storms. Details concerning these stormsand their dominant updrafts (deduced from radar reflectivity and Doppler velocity divergence signatures atupdraft summits) are discussed in sections 5 and 6.Section 7 is a concluding discussion.2. Characteristics of splitting thunderstorms The largest documented outbreak of splitting stormsoccurred on 25 August 1965, when seven splits tookplace during the afternoon in east-central Iowa withseveral of the storms moving into Illinois (Achtemeier1969). However, the most unusual documented splitting process occurred on 3 April 1964 in northwestTexas and southwest Oklahoma (e.g., Wilk 1966; Fujita and Grandoso 1968; Charba and Sasaki 1971 ). Notonly did an isolated storm split but both the right moverand left mover split, and the left-mover's right moversplit produced a third-generation pair of divergingstorms. One of the second-generation right movers produced the devastating Wichita Falls tornado. This splitting storm event was modeled numerically by Wilhelmson and Klemp (1981). Burgess et al. (1976) tabulated the basic characteristics of 16 splitting storm pairs documented in the literature; this tabulation has been updated to 31 pairs andis listed in the appendix. Table 1 summarizes some ofthe characteristics of these right-moving and left-moving storms. On average, storms that moved to the rightof the mean wind (typically density- or pressureweighted mean from cloud base to the tropopause) deviated 29- at a speed of 11 m s-~. Left-moving stormsdeviated an average of 26- at 17 m s- ~. In fact, the leftmoving member of every splitting storm pair in TableA I moved faster than its right-moving counterpart.Both right movers and !eft movers produced hail, butonly the right movers consistently produced tornadoes.Most storms probably produced damaging winds, butthat fact most likely was overlooked by storm reportersbecause hail or tornado damage was more noteworthy. Evolution of the radar reflectivity patterns during andfollowing the splitting process is quite predictable.Achtemeier (1969) likened the process to mitosis andcytokinesis observed in biological cells. He proposedthe following sequence of events for storms, as amplified by Burgess et al. (1976): 1 ) Formation stage. A thunderstorm develops andpropagates generally eastward or northeastward, notnecessarily in the direction of the mean wind. A pronounced radar reflectivity gradient appears along thestorm's rear flank. 2) Elongation stage. The thunderstorm elongates toan elliptical shape and the major axis of the ellipse isgenerally perpendicular to the direction of storm motion. During this stage, splitting of the intense reflectivity core is observed. The "split cores" grow apartTABLE 1. Statistics of post-split severe thunderstorms based ontabulations in the appendix. Standard deviations are in parentheses.Parameters Right mover Left moverAverage deviation from mean wind 29- (20-) 26- (15-)Average storm speed (m s ~) 11 (4) 17 (5)Average storm duration (h) 2-3 2Severe weather: hail (--> 1.9 cm) 65% 50% damaging wind > 17% >7% tornado 40% 10%Number of storms 31 31and the reflectivity gradients intensify along the left andright storm flanks. 3) Splitting stage. The central portion of the echorapidly diminishes in size and intensity. This dissipation leaves two separate thunderstorm cells. 4) Deviation stage. To a viewer looking along thedirection of motion, the left member moves noticeablyto the left of the mean wind direction and increases inspeed. Conversely, the right member moves noticeablyto the right of the mean wind and decreases in speed.Owing to the deviate motion, a considerable horizontaldistance develops between the two echoes. The relatively few Doppler radar studies of splittingthunderstorms (e.g., Brown et al. 1973; Burgess et al.1976; Bluestein and Sohl 1979; Burgess 1981; Conwayand Weisman 1988; Kubesh et al. 1988; Bluestein andWoodall 1990; Brown. 1992) indicate that it is typicalfor the updraft in the right-moving storm to be rotatingcyclonically and the updraft in the left-moving stormto be rotating anticyclonically. However, in some ofthe cited cases, the vorticity collocated with the updraftwas not strong enough to have it categorized as amesocyclone. Nominal threshold vorticity values formesocyclones are a function of radar resolution andvary from roughly 1 x 10-2 s-~ near the radar to 5x 10-3 s-t at distances of 200 km or more (e.g.,Vasiloff et al. 1993).3. Data sources and analysis techniques The data used in this study were obtained during theNorth Dakota Thunderstorm Project that took placefrom 12 June through 22 July 1989 in south-centralNorth Dakota. The locations of the data sensors relativeto the Standing Rock storms are shown in Fig. 1. TheCross-chain Loran Atmospheric Sounding System(CLASS) rawinsonde data and the Portable AutomatedMesonet (PAM) surface data from the National Centerfor Atmospheric Research (NCAR) were collected atElgin and Goodrich. National Weather Service (NWS)surface and rawinsonde data were collected at Bismarck. NCAR' s CP-3 and CP-4 Doppler radars were locatedin Bismarck and near Center, North Dakota, respec2054MONTHLY WEATHER REVIEWVOLUME 1221989 Nodh DakotaThunderstorm Project27 June 1989 ~Standing Rock Storms 1,2, 3200ElginCP-3BismarckGooddchI Standing RockI Indian ReservationI 0 50 km FIG. 1. The North Dakota Thunderstorm Project area during Juneand July 1989. The areas traversed by Standing Rock storms l, 2,and 3 are indicated by shaded ovals. The region within the two circles, excluding the small overlap region between the radars, is thedual-Doppler coverage area where angles between the radar beamsfrom CP-3 and CP-4 radars intersect at 20- or more and 160- or less.Rawinsonde and surface observational sites were located at Goodrich, Bismarck, and Elgin.tively. Though all or most of each Standing Rock stormwas within the nominal dual-Doppler coverage area(between-beam angles greater than or equal to 20-),storms 1 and 2 were more than 120 km from CP-4,making the data too coarse for good dual-Doppler computations. (The lowest data were 2 km above theground, and the radar beamwidth was greater than 2km.) Dual-Doppler data for these two storms were obtained from the National Oceanic and Atmospheric Administration (NOAA) WP-3D airborne Doppler radarand the ground-based CP-3 Doppler radar. Table 2gives the characteristics of both radars. Since the airborne Doppler radar scans in a verticalplane normal to the aircraft's fuselage, ideal dualDoppler geometry occurs when the aircraft flies in aradial direction directly toward or away from theground-based radar. Between 1820 and 1926 CST 27June, the WP-3D aircraft flew around and between thetwo splitting storms of interest (Fig. 2). On eight flightlegs, the airborne radar was oriented 400-90- from theradial viewing direction of the ground-based radar.Dual-Doppler analyses were performed at the following common reference times: 1822, 1833, 1837, 1844,1854, 1905, 1920, and 1924 CST. The velocity data from each radar were edited to remove noticeably bad values and to dealias velocity values. This was done using a combination of auto matic dealiasing (Eilts and Smith 1990) and user-in teractive dealiasing and error checks. When radar data are interpolated to a common ref erence time, it is necessary to assume that the reflectivTABLE 2. Characteristics of NOAA airborne WP-3D and NCAR ground-based CP-3 Doppler radars. WP-3D CP-3 radar radarTransmitterFrequency (MHz) 9315 _+ 11.6 5500Wavelength (ca) 3.22 5.45Pulse width (/~s) 0.5 1.0Pulse repetition frequency (Hz) 1600 750-1667Peak power (kW) 60 400/1000Receiver Dynamic range (dB) 80 100AntennaBeamwidth--horizontal (deg) 1.35 1.02Beamwidth-- vertical (deg) 1.9 1.02Gain, main beam (dB) 40 42.6First sidelobe (dB) -23 -23Rotation rate (deg s ~) 48 0-25 azimuth 0-15 elevationFirst trip range (km) 76 90-200Nyquist velocity (m s ~) 13 10-23ity and three-dimensional flow fields are not changingduring the time it takes the radar to make a full threedimensional scan of the storm. To help satisfy that assumption, data collection should be fast enough so thatthere is not appreciable evolution of the fields duringthe scanning time. The CP-3 radar typically scanned a FiG. 2. Flight path of the Wp-3D aircraft in the vicinity of StandingRock storms 1 and 2. Times along the path are CST; tick marks areat 2-min intervals. The polar grid indicates range and azimuth fromthe CP-3 radar in Bismarck.SEIrFEMBER 1994 BROWN AND MEITiN 2055storm in 3 min, and the P-3 airborne radar took 3-6min, so the fields should not have changed much duringdual-Doppler data collection times. The edited radar data then were interpolated to athree-dimensional grid array where dual-Doppler analyses were performed (Jorgensen et al. 1983). A motionvector of 5 m s-l from 270- was used to spatially adjustthe data to a common reference time. The resulting horizontal wind vectors were again edited interactively toremove any vectors that appeared to be erroneous; erroneous values often were found near the flight track. Since the dual-Doppler analysis technique requiresquasi-horizontal Doppler velocity measurements, onlyaircraft data within 60- above and below the flight altitude were used. Therefore, to sample an entire storm,the aircraft ideally should be at an altitude midway between the ground and storm top and at least as far awayfrom the storm as storm top is above the aircraft. Onthis day, the aircraft was flying low (2.5 km aboveground level) and often was closer than 10 km fromone or both of the storms. As a result, the updrafts eitherwere only partially sampled or were not sampled at allbecause they were completely outside the dual-Dopplercoverage area. However, dual-Doppler data did permitthe reconstruction of the flow fields downwind of theupdrafts. To compensate for the loss of information about thestrength and location of updrafts, single-Doppler velocity and reflectivity signatures from the ground-basedCP-3 radar were used to estimate updraft location andrelative strength. This dataset (consisting of 43 volumescans) was independently processed and analyzed using the multiple,Doppler radar analysis (RADAN) system (e.g., Brown et al. 1981). The primary Doppler velocity signature used in thisstudy was that for divergence near the top of the updraft(e.g., Brown and Wood 1991 ). Occasionally, there wassufficient radar return at low altitudes to detect the signature of convergent flow into the lower portion of theupdraft. The reflectivity signature used for an updraftwas the top of a rising maximum in the horizontal reflectivity field produced by hydrometeors that formedin the upper portions of the updraft and that were carried upward by the vertically growing current.4. Environmental situation The research area of the North Dakota ThunderstormProject in south-central North Dakota was under theinfluence of a 500-mb ridge on 27 June 1989. At 1800CST, the 500-mb high was centered over northernMexico. The ridge extended sharply northward, and theridge axis was lying just to the west of Bismarck. On the morning of 27 June, a very weak warm frontextended from a surface low in Alberta through westcentral Montana and across central South Dakota.There was not much temperature contrast across thefront but the air south of the front was considerablymore moist. The front moved northward during the dayand by early afternoon was along the northern borderof South Dakota. The satellite photos in Fig. 3 showthe evolution (at 60-75-min intervals) of a northsouth line of thunderstorms in South Dakota and northern Nebraska that formed in the moist air south of thefront. The storms of interest in this study developed atthe northern end of this line and are identified inFigs. 3c,d. The dewpoint temperature and wind plots in Fig. 4document the northward progression of the front andthe moist air behind it from 1800 CST 26 June through1800 CST 28 June. The stations in the plot are alignedfrom Elgin to the southwest through Bismarck to Goodrich to the northeast (refer to Fig. 1 ). The plots showan increase in dewpoint temperature during the morning of 27 June. Dewpoint temperatures of about 15-Cat Elgin during the afternoon and evening of 27 Junemarked the northern extent of the moist air that led tothe formation of thunderstorms (including StandingRock storms 1, 2, and 3) that occurred in South Dakotaand extended just across the border into North Dakota(Fig. 3). During the morning of 28 June, increasinglystrong southeasterly winds brought the moist air farthernorth into central North Dakota, where severe thunderstorms occurred in the project area during the afternoon and evening of 28 June (Boe and Johnson 1990). The atmospheric soundings closest in time and spaceto the Standing Rock storms were the CLASS rawinsonde released at Elgin at 1430 CST and the operationalNWS rawinsonde released at Bismarck at 1800 CST27 June, both of which were ahead of the warm front.The Elgin sounding (Fig. 5) reveals a lapse rate that isnearly dry adiabatic from the surface to 600 mb, whichis favorable for convection. The relatively dry airshown at the surface in Fig. 4 extends upward throughmost of the sounding. The Elgin soundings at 0830, 0923, and 1130 CSTthe next morning (not shown) indicate that temperatures below 600 mb and above the eroding nocturnalsurface inversion had increased by 2--3-C followingwarm front passage. Moisture between the surface and850 mb was steadily increasing, as also reflected by thesurface measurements in Fig. 4. The layer from 850 to700 mb was drier than indicated in Fig. 5. On the otherhand, the layer from 700 to 300 mb was moister, probably reflecting residual moisture from the storms on 27June. A composite hodograph was prepared from the 1430CST Elgin and 1800 CST Bismarck soundings bysmoothing the average u and v components (Fig. 6).The density-weighted mean wind in the 1 - 12-km layerfor the composite hodograph is 268- at 6.5 m s-~.Overall, the hodograph is essentially a straight line, thetype that numerical models indicate should produce asymmetric pair of splitting storms (e.g., Rotunno and2056 MONTHLY WEATHER REVIEW VOLUME 122FIG. 3. Visual GOES-7 satellite photographs of cloud evolution over the Dakotas area at (a) 1501, (b) 1601, (c) 1716, and (d) 1831 CST27 June 1989. Standing Rock storms I and 2 are identified with arrows; L and R indicate the left- and right-moving members of storm 1.Klemp 1982). There is some curvature to the hodograph in the lowest 3 km that suggests that the rightmoving member of a splitting storm pair should bemore dominant than the left-moving member. According to Brown (1993), the nearly uniform wind layerfrom 3 to 8 km is the type found within, but not necessarily limited to, supercell thunderstorm environments. Davies-Jones et al. (1990) proposed two empiricalcriteria for mesocyclone (rotating updraft) formation,namely, that the storm-relative winds in the lowest 3km should be at least 10 m s-~ and that they shouldveer by at least 90- through the same depth. Windsrelative to Standing Rock storms 1R and 2R in Fig. 6satisfy only one or the other of these criteria. Likewise,winds relative to left-moving storms 1L and 2L satisfyonly the 90- change of direction criterion for mesoanticyclone formation (where veering winds are replacedby backing winds). Therefore, theory indicates that theupdrafts in storms 1 and 2 may not develop significantrotation. Since the composite hodograph in Fig. 6 representsconditions north of the Standing Rock storms and theother preexisting thunderstorms, it properly can be usedonly to deduce flow toward the left flanks of storms 1Land 2L. Air approaching the right flanks of storms 1Rand 2R likely was modified by the low-altitude outflowfrom the nearby storms to the southeast and east.5. Storm evolution Standing Rock storms 1 and 2 were the stormswithin which dual-Doppler data were collected usingthe airborne and ground-based radars. Of the eightdual-Doppler analysis times, the wind and reflectivity fields at 7-km height for six nearly equally spacedtimes are shown in Fig. 7. At the earliest analysistime (1822), storm 1 shows an elongation of the 20SEPTEMBER 1994 BROWN AND MEITJN 2057GoodrichBismarck Elgin I I ~.s ~ I %% ~,, ~I /i !~ ~,~--~!!512.5 [ ~lZSI I I I I I Wind Speed (m s'1) /5 ,is lo, ? I ! I [~/ ~,~ J ',..5 X I Wind Vector .I-'J,U,." ....... ' ...... Elgin 18 O0 06 12 18 O0 06 12 18 28 June 27 June 1989 28 June 1989 Time, CST FIG. 4. Space-time cross sections of dewpoint temperature (-C),wind speed (m s-~), and wind vectors (long barb is 5 m s ~) from1800 CST 26 June through 1800 CST 28 June 1989. As can be seenin Fig. 1, the cross section extends in a southwest to northeast direction from Elgin through Bismarck to Goodrich.dBZ contour normal to the mean wind direction, asis typical of storms that are starting to split (e.g.,Achtemeier 1969). The dual-Doppler coverage areadid not reach the main portion of storm l, becausethe aircraft turned toward the perpendicular to theviewing direction of the CP-3 radar. Though thewestern portion of storm 2 has missing winds owingto the proximity to the aircraft track, the winds in therest of the storm seem to indicate that the midaltitudeflow was around the storm's core. Stronger winds areindicated along the north and south edges of thestorm, and weaker winds are evident in the wake region downstream of the storm's center. By 1833 CST, the midaltitude reflectivity field associated with storm I had split into two identifiableentities (labeled 1L for the left member and 1R for theright member). Air downstream of the left membershows confluence into the wake region. With time,storms 1L and 1R moved farther apart and, unlike thetypical splitting storm, storm 1L was the larger andlonger-lasting member of the pair. Though not complete, dual-Doppler coverage ofstorm 2 was more consistent. As the storm started tosplit at 1844 CST, enhanced midaltitude flow wasmaintained on the left side of the left member and onthe fight side of the right member. The wind speed100 ,?500[--' ...........~oo-I /,k~~~, ~'~~~ ',~CF ~L~1050/ / ~ ~ '~ Y I ~l r ~ ~ Elgin 27 June 1989 1430 CSTg FIG. 5. Temperature, dewpoint temperature, and winds from thesounding released at Elgin at 1430 CST 27 June 1989. The long windbarb is 5 m s-~ and the flag is 25 m s-~. Heights are in kilometersabove sea level.minimum between the two members was maintainedduring the splitting process through 1854 and was stillsomewhat apparent at 1905 CST. Throughout the splitting process, the cross-stream variation of the horizontal wind (that is, wind speed minimum in the middleand maxima on the left and right sides) suggested thepresence of cyclonic vertical vorticity (where windspeed decreases from right to left) associated with the180- .10 COMPOSITE HODOGRAPH - 2L 27 JUNE 1989 2 olL '~""~----~ 37 ..1R3 20 30 2R ' -~9 ~0'6 ' '"'~1" 1 2 km: 2700 5 -10 rn s43600 Fla. 6. Composite hodograph computed from the 1430 CST Elginand 1800 CST Bismarck rawinsonde releases on 27 June 1989. Arrowis the mean wind vector in the l -12-km layer. Dots indicate the tipsof the storm motion vectors for the left (L) and right (R) portionsof storms 1 and 2 plotted relative to the hodograph origin; stormmotion was determined from the locations of the peak heights ofsequential updrafts within each storm.2058 MONTHLY WEATHER REVIEW VOLUME 122-65-70-75-80-85-90 -9."-10C-105-11~o11~ i I ~ ~ i I ~ ~ i 7 km 1822 CST 27 June 1989 ~ ~/-~ ~ '~ ~ ~ ~ I ~ I ~ I I I ~-70 -55 -C~ -55 -50 -45-aO -g5 -aO -25 -20 7 km 1833 CST-7o 27 June 1989-75-85-100-11o/-115 I I I I I [ I I I -70 -65 -60 -55 -50 -45 -40 -35 -30 -25 -20 - ~ '65/ I I ~ . I ~ I ~ I 6. I i ' ' ' I ' ' ' ' / -70Fl -75 -75 -85 J;'-:: _2 ~~ t -110 7kin ~ / 7kin ~ 27 June 1989 18~ CST1 -110~ 27 June 1989 1854 CS~ 115 I I I I I I I I I / _115/ I I I I I i i [ I - -70 -65 -~ -55 -~ -45 -40 -35 -30 -25 -20 -70 -65 -60 -55 -50 -45 -40 -35 -30 -25 -20 -65- -65~ , [ ~., ,- ~ ~ , ~ , -75 - -75 -80- -80 x - ~ -90- -90~ ~ ~105 ~ I / 7 km I / 7 km -110- 27 June 1989 1905 CST~ -110~ 27 June 1989 1920 CS~ -115 -~5 -40 -35 -30 -25 -20 -70 -65 -60 -55 -50 -45 -40 -35 -30 -25 -20 -70 -65 -60 -55 -50 East - West dis~nce (kin) from CP3 F~G. 7. Rad~ ~cQcctivit~ contours (20, 30, and 40 dB~) and ~ad~-rclative horizontal wind vectorsthcsized f~om dual-Dopple~ rad~ measurements) at a height of ? ~ above the ~rou~d at reference times(a) 1822, (b) ]833, (c) ]8~, (d) ]854, (c) ~905, and (f) 1920 CST 27 June 1989. ~cn~th of vecto~ atuppc~-d~ht comer ~cpresents l0 m s-~. ~c potion of the WP-3D Qi~ht ~ack used in the dual-Doppleranalysis is i~dicatcd.SEPTEMBER 1994 BROWN AND MEITiN 2059-901735 CST 10 km 1740 le 1E~ % ld dBZ ld lc~'~ lc ~a b I I I I 1745 1750 10 le 10 ldc d I I I IE . -100d.O -llC,~ -@0O~~ -100OO)I~ -110O -120 -60 -50 -40 -30 -50 -40 -30 EAST-WEST DISTANCE FROM CP-3, kmFIG. 8. Example of evolution of radar reflectivity features at 10-kmheight associated with the evolution of deduced updrafts Ic-le.right member and anticyclonic vorticity (where windspeed decreases from left to fight) associated with theleft member. Based on the limited number of singleand dual-Doppler radar studies of splitting storms,these deduced vorticity fields are consistent with theobservations of cyclonically rotating updrafts in rightmoving storms and anticyclonically rotating updraftsin left-moving storms. However, detailed single-Doppler radar observations discussed in the next section donot give any indication of updraft rotation in eithersplitting storm.6. Deduced updraft evolutiona. Single-Doppler radar signatures of updrafts Updraft positions had to be estimated using the locations of peak reflectivity values and single-Dopplervelocity divergence signatures near updraft summits,because (a) neither updraft magnitude nor evolutionwithin storms 1 and 2 was available from the airborneWP-3D and ground-based CP-3 Doppler radars owingto inadequate dual-Doppler' coverage, and (b) thestorms were too distant from the CP-4 Doppler radarto provide acceptable spatial resolution for dual-Doppler coverage by CP-3 and CP-4. Kingsmill and Wakimoto ( 1991 ) showed, in their multiple-Doppler radarstudy of a severe thunderstorm developing within aweakly sheared environment, that updraft and reflectivity summits typically were within a kilometer of eachother and were associated with the rising turret of thevisible cloud. In this study, the single-Doppler radardata collected by the ground-based CP-3 radar at 3-5min intervals provided sufficient temporal and spatialresolution to deduce the evolution and relativestrengths of the updrafts. With rising reflectivity maxima being used to indicate the locations of updraft summits, the following terms are used interchangeably: reflectivity maximum, reflectivity signature of updraft,deduced updraft, and updraft. The typical evolution of successive updrafts as indicated by radar reflectivity patterns is illustrated inFig. 8 for storm 1. Although it might appear that therewas a single updraft with small-scale perturbations,close examination of the reflectivity data in four dimensions reveals a series of updrafts that form successively on the storm's left flank. At 1735 CST, reflectivity maximum lc was just starting to descend afterreaching its maximum height. At successive 5-min intervals, it changed from a reflectivity peak to a rightflank protrusion as it moved downstream (toward theeast-southeast) and continued to descend. MaximumI d reached its greatest height at 1740 CST and likewisechanged into a protrusion as it descended and moveddownstream. Reflectivity maximum le made its initialappearance as a left-flank protrusion and continued togrow in intensity and height until reaching maximumvertical extent just after 1750 CST. It was common,especially in storm 1, for the reflectivity signature tomove very slowly as it was growing vertically, but afterit started to descend it would increase speed and movedownstream basically with the ambient flow. Since the divergence signature near storm top provides a direct indication of updraft location, it was usedto verify the reflectivity estimate of the center of theupdraft whenever it was present. At the height wherethe divergence signature was most pronounced (typically a few kilometers below storm top), a comparisonwas made between the center of the signature and thecorresponding center of the reflectivity maximum foreach updraft at each reference time in both storms 1and 2. The distribution of reflectivity signature centersrelative to the corresponding divergence signature centers is shown in Fig. 9. Of the 76 comparisons, 21 ofthe reflectivity maxima were collocated with the centerof the divergence signature. A total of 66 reflectivitymaxima were within 2 km of the corresponding divergence center and the remaining 10 were located 2-4km to the east through south. If precipitation particles(represented by a reflectivity maximum) were carrieda few kilometers downwind relative to the updraft, thehodograph in Fig. 6 indicates that they would moveslightly south of east. Interpretation of the horizontal flow fields within the storms is based on single-Doppler velocity measure ments. Since each single-Doppler velocity data point is a one-dimensional (that is, radial component) repre sentation of the three-dimensional flow field at that lo2060MONTHLY WEATHER REVIEWVOLUME 122cation in space, there is no unique way to reconstructthe three-dimensional velocity vector at that location.However, when presented in a quasi-horizontal twodimensional display, there are a number of assumptionsthat can be made to reconstruct a plausible two-dimensional flow field (e.g., Lemon et al. 1978). Also, thereare singleTDoppler velocity signatures that can be interpreted to represent specific horizontal flow patterns(e.g., D-naldson 1970; Brown and Wood i983, 1991 ).Some of' these signatures within a thunderstorm represent axisymmetric convergence at the bottom of anupdraft (top of a downdraft), axiSymmetric divergenceat the top of an updraft (bottom Of a downdraft), axisymmetric rotation within an updraft or downdraft, anda pair of counterrotating axisymmetric vortices (rotating about vertical axes) on the midaltitude flanks of astrong updraft. These types Of signatures form the 'basisfor issuing severe thunderstorm warnings based on single-Doppler velocity observations~ An example of the, vertical distribution of singleDoppler velocity .signatures associated with a nonrotating updraft is shown in Fig. 10-for updraft lg instorm l'at '1818 CST~ The figure shows single-Dopplervelocity' signatures Of convergence, a vortex pair, anddivergence at 1, 7, and 12 km, respectively. Superimposed on each Signature is a hypothetical flow field thatcould result When a representative Doppler velocityvalue (for example, 3, 5, and 8 m s-d, respectively)was added to the signature at each height to make thesignature more' ibalanced with zero Doppler velocity atthe center; this process produces a feature-relative flowfield at each height. The convergence signature at 1: kmshows a maximum of flow away from the radar (+ 1m s-1.) on the near side(north.side) of the signatureand a maximum of flow toward the 'radar (-7 m s-l)on the far side (South side). At 12-km height, the OPposite is true. The divergence signature shows a maximum of flow toward the radar (~-.! 5 m s - l ) on the nearside and a minimum of flow toward the radar (-1m s- 1 ) On the far side. Note that the center of the 25dBZ contour 'at 12 km was about 1 km to the eastsoutheast of the center of' the divergence signature. The S!ngie-Doppler .velocity signature at 7 km forthe vortex pair is more complicated. When the radar isviewing the updraft essentially perpendicular to the environmental flow direction, as in this case, the signatureideally is like a four-leaf clover of alternating' sign[ compare with the idealized 280- panels in Figs. 18 and19 Of Brown and Wood (1991)]. On the near side(north side) of the updraft, flow is toward the radar(negative extreme), and then away from the radar(positive extreme) as the ambient air flows around theupdraft. If the updraft were centered on this' negativepositive couplet, the signature would be called a "mesoanticyclone signature." On the far side (south side),flow is away from the radar (positive extreme), andthen tOWard the radar (negative extreme). If the updraftwere centered on this positive~negative couplet, thesignature would be called a "mes0cyclone signature."Since the centers of the upper-altitude divergence andreflectivity signatures are nearly vertically aligned withthe center Of the col between the two pairs of positiveand negative extremes, one deduces that the midaltitude updraft Was located between the two v0rticity centers and therefore had insignificant net rotation. If thereflectivity signature had been in error by an extremeof 2'4 km, the error would have been in a southeasterlydirection (Fig. 9), which is normal to the directionneeded to move the updraft from the center of' the colto the center of cyclonic or anticyclonic rotation. Therelative relationships shown in Fig. 9 are similar for allUpdrafts in these storms. Therefore, use of the Centerof the upper-altitude reflectivity maximum as an indicator of updraft location (center of divergence signature) Should not lead to problems in discriminating between rotating and nonrotating updrafts.b. Deduced updraft features in storms 1 and 2 During the growing stage of an updraft, the hydrometeors that form in its upper portions produce a localized maximum of radar reflectivity in plan view..The15-dBZ top of this feature was used here to approximate the summit of each updraft within storms 1 and2, As long as the 15-dBZ top was rising with time (correspOnding to Visual cloud turrets), one can' deduce thatthe updraft was intensifying and increasing in altitud&When the 15'dBZ top reached it's maximum verticalN~ o NW / %~~1-3WSWI 10-15 16+4 km FIG. 9. Distribution of the location of the reflectivity signature ofan updraft center near storm top relative to the center of the singleDoppler velocity signature of the associated divergence center. Ofthe 76 distinctive refiectiVity-divergence pairs in storms I and 2, 66were within 2 km of each other, including 21 that were coincident(black circle at the center of the distribution).SEPTEMBER 1994 BROWN AND MEIT[N 2061/ / / r1 km 6-27-897 km 1818 CST12 km .8;/ -:4 ~'. ~4- d*:/-4~, 0,~ $q FIG. 10. Single-Doppler velocity measurements (radar-relative) at1-, 7-, and 12-km heights within storm 1L in the vicinity of updraftlg at 1818 CST. Superimposed streamlines represent idealized feature-relative flow. The thick outer boundaries are 5-dBZ reflectivitycontours, and the small thick inner closed curve at 12 km is a 25dBZ contour. Border tick marks are at 5-km intervals. Arrows outsideeach panel indicate the CP-3 Doppler radar viewing direction passingthrough the center of each panel.extent and started to descend, one then can deduce thatthe updraft had started to die and that the radar-detectable particles probably still represent the summit of theweakening updraft. However, after a short time, thedescending top no longer can be interpreted as representing an updraft, but instead should be interpreted asrepresenting the upper portion of the descending precipitation column within the storm cloud. The reflectivity maximum associated with hydrometeors thathave been descending downstream of the updraft for10 or more minutes typically has spatial continuitydown to the ground. A time history of the tops of the primary 15-dBZreflectivity features in Standing Rock storm 1 is shownin Fig. 11; to be called a primary feature, it generallywas the tallest feature in the storm sometime during itsexistence. The plot in Fig. 11 indicates that successivededuced updraft summits grew to increasing heightsuntil updraft I e, which had the tallest and longest-lasting reflectivity feature. Subsequent updrafts grew tolesser and lesser heights until the storm dissipated withthe demise of updraft lj. Updrafts lc-lh penetratedthe tropopause by 1-4 km. The reflectivity signatureof the first updraft on the storm's right flank (1Ra)appeared 20-25 min after le reached its maximum vertical extent. The strongest right-flank updraft was thesecond of the four to occur, but none of them werestrong enough to penetrate the tropopause. None of the updrafts in the left or fight portions ofstorm 1 ever exhibited any significant rotation. Significant rotation (mesocyclone) is defined as the collocation of an updraft with a region of Doppler velocityazimuthal shear exceeding 5 x 10-3 s-l or vorticityexceeding 1 x 10-2 s-i; for an axisymmetric vortexwith a Rankine velocity profile, vorticity is equal totwice the azimuthal shear of Doppler velocity acrossthe single-Doppler velocity vortex signature (e.g.,Brown and Wood 1991 ). However, nearly all the taller(and, presumably, stronger) deduced updrafts were located within a vortex-pair signature at midaltitudes.Some of the new updrafts formed within one of themembers of the vortex pair and briefly were coincidentwith weak azimuthal shear, but when they becamestrong they had their own vortex-pair signature. As thereflectivity signatures were descending, a few of themwere collocated with weak azimuthal shear of one signor the other. The hatched portions of the curves in Fig. 11 represent the times when a midaltitude vortex-pair signature (indicative of flow around a strong updraft), coincided with an updraft signature. Of those updrafts thathad 15-dBZ tops that penetrated the tropopause, onlyib and lh did not have an associated vortex-pair signature; these generally were the more short-lived reflectivity features. Of the updraft signatures that did notreach the tropopause, only 1Rb had an a, ssociated Vortex-pair signature. The horizontal projections of the 15-dBZ tops in Fig.11 are plotted in Fig. 12. These features (representingupdrafts and then downdrafts) moved predominantlyfrom west to east with the mean wind. The letter oneach line indicates the location of maximum verticalextent for that updraft signature. It is apparent' fromFigs. 11 and 12 that each 15-dBZ top remained essentially stationary during !ts growth phase and thenmoved downwind following updraft demise as the reflectivity maximum descended in Conjunction' with thedescent of precipitation particles.18E 16 12>' l041700STORM I d ~ f g , , , , , t ,t - 1730 1800 1830 TIME, CST27 JUNE 1989 Left Portion .... Right Portion ~ Vortex PairI S!gnat~ro ?,~ ? ~t ,? 1900 1930 FIG.. I 1. Time-height plots of the 15-dBZ tops of radar refle~tivityfeatures associated with deduced updrafts in storm 1. Solid curvesrepresent features in the initial left member and dashed cur~es represent features in the right member. Hatched portions of the curvesindicate the times when there were single-Doppler velocity signaturesof vortex pairs at midaltitudes (indicating horizontal flow aroundboth sides of the updraft). Vertical arrows along the abscissa indicatethe times of the dual-Doppler data presented in Fig. 7. Tropopauseheight (dashed) is a representative height derived from the'varioussoundings made on 27 June.2062 MONTHLY WEATHER REVIEW VOLUME 122 EZO-70-80-901/1L27 JUNE 19891847 ~i~o533~01T,~ 40-100 3O 20~-110 10~.~.~5 1802 1710.43. 13~58 ~1 4g- -- 33~806"'-'~ 33-lOO-1101R1908 I I I I-60 -50 -40 -30 -20 -10 EAST-WEST DISTANCE FROM CP-3, km FIG. 12, Horizontal projections of the refiectivity signatures of updrafts shown in Fig. 11, with beginning and ending times indicated.The letter identifying each deduced updraft is placed at that pointalong the curve where the reflectivity signature had its maximumvertical extent.1412I08641800STORM 2 bc f27 JUNE 1989 TropopauseA Left Portion.... Right Portion \....................... Vortex Pair Signature ~t ~t t~ t ~ t ~ ? ~ 1830 1900 1930 TIME, CSTFIG. 13. Same as Fig. 11 except for storm 2.2OOOa few tens of kilometers to their east and southeast (seeFig. 3 ). However, the outflow boundaries from storms1 and 2 that triggered the series of new left-flank andright-flank updrafts were not detectable in the radardata. By comparing the initial location of new updraftswith the edge of the storm, one can surmise the locationof the triggering gust front. The initial location aloft ofthe reflectivity signature for each new updraft relative-60 E The horizontal projections in Fig. 12 indicate that e~ -70storm 1 / 1L propagated to the left from the very begin- &ning, because new updrafts consistently formed on the oleft flank. Storm 1R propagated more in the downwind ~ O -80direction rather than exhibiting pronounced propaga- ec ,,tion to the right of the mean wind. tu Vertical and horizontal evolution of the 15-dBZ tops o Zof the reflectivity features in storm 2 is shown in Figs. < -9013 and 14. The overall evolution was the same as for o~storm 1. The tallest reflectivity feature was associated ~with updraft 2c. About 10 min after updraft 2c reached ~ -100its maximum vertical extent, a new updraft formed on = Othe storm's right flank. In all, three significant updrafts o~appeared in storm 2R, and the first one was the stron- -I F- -90gest. Each of them formed at essentially the same ~:ground-relative location. All updraft signatures within ~ostorms 2/2L and 2R exceeded 10 km in height, and allof them, except for updraft 2b, coincided with a -100midaltitude vortex-pair signature. None of the updraftsin these storms exhibited any significant rotation.The evolution of storms 1 and 2 on the radar display(not shown) suggests that both storms formed along anoutflow boundary that spread out from thunderstorms2/2L 27 JUNE 19891934 ~ 49 24i ~42 13% 41901-~ 44Ld,,..--~ 54 40/22 ~C--'--'--- 5018062RI I I I1844fa,..F~19o1 ~7~...~.-38 1908.---~'~'~.~.~..24 I I I I-80 -70 -60 -50 -40 -30 EAST-WEST DISTANCE FROM CP'3, km F~G. 14. Same as Fig. 12 except for storm 2.SEPTEMBER 1994 BROWN AND MEITiN 2063to the 20-dBZ radar reflectivity contour at 1-km height(approximating the edge of the storm) in storms 1/ILand 2/2L is shown in the top parts of Figs. 15 and 16.The figures indicate that each new updraft signature inthe initial and subsequent left portions of the stormsformed above or within a few kilometers of the leftedge of the storms. Similarly, the updraft signatureswithin storms 1R and 2R formed within a few kilometers of the right edge of the storms (bottom part ofFigs. 15 and 16). These data suggest that the formationof the right member of each splitting storm was delayeduntil after the subcloud cold pool had built up enoughfor the gust front to reach the right side of the storm.7. Concluding discussion-60-70-80-90 ~ -100During the late afternoon and early evening of 27 o ~nJune 1989, thunderstorms formed over Standing Rock 4Indian Reservation in the southern portion of the North ~ eeDakota Thunderstorm Project area. Two of at least othree splitting storms on that day were investigated us- :~ing the NOAA WP-3D airborne Doppler radar and theNCAR CP-3 ground-based Doppler radar. The result_801 1/1L 2/2L ~f~ 1 km, 20 'dBZ(~ ~ ~ i J g f 27 JUNE ~1989 ~ ~-70 -60 -50-40 EAST-WEST DISTANCE FROM CP-3, km FiG. 16. Same as Fig. 15 except for storm 2.-80-90 -100-30 -9o-100-110-lOO-110 -60 -50 -40 -30 -20 EAST-WEST DISTANCE FROM CP-3, km FiG. 15. Plan view of the initial location of each deduced updraftin the upper portions of storm 1 (circled letter) and the coincident20-dBZ reflectivity contour at 1-km height (labeled with the updraftletter).ing dual-Doppler coverage of the actively growing portions of the storms was limited owing to the aircraft'slow altitude and proximity to the storms. Both storms 1 and 2 exhibited the same characteristics during the splitting process. Each started as a single radar echo with midaltitude air flowing around it;increased wind speed was found on the lateral flanksand decreased speed was found on the downwind side.With time, the radar reflectivity pattern elongated alongan axis normal to the predominant midaltitude winddirection. As the elongation process continued, the reflectivity pattern separated into two distinct echoes thatcontinued to move apart relative to one another. Theleft members moved at 7-8 m s -~ at an angle of about65- to the left of the mean wind, while the right members moved generally with the mean wind but in a morevariable manner. Deduced updrafts within both members of the two splitting storms moved essentially inthe direction of the mean wind. Contrary to the usualsplitting storm evolution, the left member was the dominant and longer-lasting one. Since the updraft regions of the storms were not consistently sampled with the dual-Doppler radar system,radar measurements from the ground-based CP-3 radarwere used instead to deduce updraft evolution. Peakreflectivity values near storm top were used to deducethe locations of the primary updrafts; a primary updraft2064 MONTHLY WEATHER REVIEW VOLUME 122generally was defined as one that was the tallest of allreflectivity features sometime during its existence.When present, single-Doppler velocity divergence signatures near storm top were used to confirm the updraftlocations. From the very beginning of both initialstorms, each new updraft formed on the left flank of anexisting updraft. The average interval between successive updrafts was I 1 min (with a range of 3-31 min).The updrafts grew to increasingly greater heights untilone reached maximum vertical extent 30% -40% of theway through the storm's life cycle. Successive updraftsthen grew to lesser and lesser heights until the precipitation from the last left-flank updraft fell out about2 h after the storm first appeared on radar. The first right-flank updraft appeared on radar 1020 min after the tallest left-flank updraft signaturestarted to descend. The average interval between successive right-flank updrafts was 12 min (with a rangeof 4-28 min). The fight-flank storms lasted only about1 h, and their tallest tops were considerably shorter thanthose for the left-flank storms. Neither member of either splitting storm had any rotating updrafts, contrary to what typically has beenfound from Doppler velocity measurements in a limitednumber of splitting storms. Of the left- and fight-flankupdrafts in storms 1 and 2 that extended upward beyond 10 km, 80% produced single-Doppler velocitysignatures of a vortex pair, with the vertical vorticitycenters on the left and fight sides of the updrafts. Thesignatures typically were limited to the height intervalfrom 6 to 8 km and never extended below 5 km orabove 9 km. Vorticity values associated with the vortex-pair signatures typically were _+2 x 10-3-4x 10-3 S-t, with a rare extreme value of _+5 x 10-36 x 10-3 s-~. It is interesting to hypothesize about what controlledthe propagational characteristics and relative timing ofthe left- and right-flank updrafts in the Standing Rockstorms. Low-altitude convergence along cold-air outflow boundaries plays an important role in the initiationand maintenance of updrafts within convective storms(e.g., Thorpe and Miller 1978; Weaver 1979; Wilhelmson and Chen 1982). Since the initial updrafts withinStanding Rock storms 1 and 2 formed on the left side,it is likely that the cold surface outflow air initially wasable to extend only beyond the left edge of the stormand interact with gust-front-relative flow approachingthe storm at low altitudes. There also may have been acontribution from cold-air outflow from the storms immediately to the east and southeast. Subsequent sequential updrafts led to the formation of precipitationdowndrafts that systematically reinvigorated the surface cold pool. It is likely that the increasing mass of.surface cold air associated with those downdrafts permitted the gust front to eventually extend to the rightedge of the storm. Outflow air from the storms to. theeast and southeast also may have retarded the low-altitude outflow toward the right flank. Gust-front-relative flow approaching the right side of the storm likelyproduced a region of low-altitude convergence that ledto the formation of the initial right-flank updraft. Theright member of the split then developed its own precipitation-cooled and evaporatively cooled air, and gustfront, and was able to maintain its own regenerationmechanism as the left member moved away. The above discussion also provides some clues towhy the left-moving storms were the dominant members of the splitting storms on 27 June 1989. Based onthe hodograph in Fig. 6, which to a first approximationis a straight hodograph, one would expect a balancedpair of splitting storms to be produced (e.g., Rotunnoand Klemp 1982). To a higher approximation, hodograph curvature in the lowest 3 km indicates that therewas greater storm-relative/gust-front-relative environmental flow toward the right flank of the right membersthan toward the left flank of the left members, eventhough both storm-relative/gust-front-relative flowsare relatively weak. If, indeed, cooler, more stable outflow from the nearby storms retarded low-altituderightward-moving outflow from storms 1 and 2, then italso could have decreased the amount of buoyant energy available for right-flank updrafts, resulting in shallower and shorter-lived right-flank storms. A dualDoppler radar study of Standing Rock storm 3, whichis under way, may provide some answers to the questions concerning the dominance of the left-movingstorms on 27 June 1989. Acknowledgments. Special thanks go to Dr. Bradley Smull of the National Severe Storms Laboratory(NSSL, Boulder) for the enormous amount of timehe spent both in training the authors on the use of theairborne Doppler radar editing programs and in consulting on problems that occurred while testing someof the programs. Thanks also to Dr. David Jorgensen(NSSL, Boulder) for his help in interpreting the airborne data. Robert Hueftle (NSSL, Boulder) washelpful in making many changes to existing programs and was always available when problems occurred. Melissa Hornecker and Kathleen Eyerman(NSSL, Norman, Oklahoma) assisted with the editing and processing of the CP-3 Doppler radar datafor the single-Doppler radar portion of this study. Weappreciate the detailed reviews of the manuscript byDrs. Erik Rasmussen (NSSL, Norman), ThomasChristian (Wave Propagation Laboratory), and ananonymous reviewer. The North Dakota Thunderstorm Project was sponsored by the NOAA FederalState Cooperative Program in Atmospheric Modification Research, the North Dakota Atmospheric Resource Board (NDARB), and the National ScienceFoundation. We especially appreciate the supportand encouragement of Project Director Bruce Boe ofthe NDARB.SEPTEMBER 1994 BROWN AND MEITiN 2065TABLE AI. Characteristics of post-split severe thunderstorms. Associated severe weather: H--haiL W--damaging wind, and T--tornado. Maximum Approximate deviation Mean wind lifetime Severe (deg) from Average speed computation Date Location Storm (h) weather mean wind (m s- ~) technique Source27 July 1956 Quebec la 2 -- 8 L 15 700 mb Hitschfeld (1960) Ib 2 -- 43 R 13 (3 km) la' I -- 29 L 17 la" I -- 8 R 1424 May 1962 Oklahoma a 3 21 L 18 900-200 mb Newton and Fankhauser B 8 H, W, T 53 R 9 (1-12 kin) (1964) pressure weight3 April 1964 Texas-Oklahoma A~ 2 34 L 13 sfc-300 mb Wilk (1966); Fujita and Cs I 21 R 11 (0-9 kin) Grandoso (1968); A2 2 10 L 16 pressure weightCharba and Sasaki C~ 2 H, W, T 15 R 11 (1971) A3 3 H 44 L 17 C3 3 H, T 19 R 13 A4 I 13 L 21 C4 2 33 R 1823 April 1964 Oklahoma D 1 H, W, T 20 L 16 850-300 mb Hammond (1967) F I 0 R 14 (1.5-9 km) pressure weight27 May 1965 Oklahoma L 2 H, T 35 L 15 storm depth Harrold (1966) R 3 H, T 35 R 1025 August 1965 Iowa - Illinois C3 3 H, W 32 R 10 Achtemeier (1969) C8 -- H 25 R 13 C9, 11 3 H 17 L 18 C17 I H 21 R 15 CI8 i H 15 L 23 El 3 H,W,T 15R 8 E2 2 H 48 L 17 E4 -- H,T 12R 13 E5 -- H 40 L 20 K1 2 H 8 L 2326 August 1965 Iowa S2 2 H 29 L 26 Achtemeier (1969) SI >3 H, T 16 R 1916 April 1967 Oklahoma A 2 H 15 L 17 950-200 mb Haglund (1969) D 3 H, T 35 R 14 (0.5-12 kin) L I -- -- pressure weight K 3 40 R 1318 June 1970 Colorado I I 0 L 17 0-10 km Marwitz (1972) 2 2 H 70 R 919 April 1972 Oklahoma L 3 H 22 L 28 storm depth Brown et al. (1973) R 3 H, T 25 R 13 pressure weight27 June 1972 Oklahoma LI 2 H, W, T 47 L 16 RI 2 25 R 10 L2 2 H 35 L 16 R2 2 22 R 1124 May 1973 Oklahoma L 2 H 28 L 15 1-12 kin R 2 H, T 12 R 10 pressure weight30 May 1976 Oklahoma L I H 54 L 14 density weight R 3 H, T 222 R* 4 I May 1977 Oklahoma L I 60 L 6 1 - 12 km R I H 54 R 3 density weight29 April 1978 Oklahoma DLM I -- 18 L 19 I - I 1 km DRM 1 -- 76 R 4 density weight KLM I -- 38 L 18 KRM I -- 56 R 5 6 June 1979 Oklahoma L 3 20 L 16 1-11 km Brown (1992) R 3 H 7 R 11 density weightI August 1981 Montana L >2 32 L 21 3 - 12 km Kubesh et al. (1988) R >5 H 7R 926 April 1984 Oklahoma L 2 22 L 24 I -10 km Burgess and Curran R 4 H, W, T I R 17 density weight (1985)13 June 1984 Colorado AI >1 13 L 9 0-6 km Conway and Weisman A 3 H 53 R 5 density weight (1988) E1 <1 2 L 10 E <1 47 R 2Achtemeier [1976, personal communication; from Burgess et al. (1976)]Burgess et al. (1976); Lemon et al. (1978)Lemon and Burgess (l 980); Burgess (1981 )Bluestein and Sohl (1979)Adams (1981, personal co~nmunication)* Storm direction is so anomalous that it is not included in Table 1 computations.2066 MONTHLY WEATHER REVIEW VOLUME 122 APPENDIX Characteristics of Splitting Thunderstorms Burgess et al. (1976) compiled a list of 16 splittingstorm pairs from the literature that itemized such stormcharacteristics as the lifetime, associated severeweather, directional deviation of storm motion from themean wind, and storm speed for both members of eachsplitting storm pair. Their compilation has been updated to include a total of 31 splitting storm pairs thatare listed in Table A1. Included in Table A1 is a listing of the methods usedto determine the mean wind, from which the deviantstorm motion was computed; not all sources mentionedthe computation method. In most cases, a weightedmean value was computed over most of the stormdepth; the overall height interval is listed in the table.To account for the decrease of air density with height,the means either were computed from values at uniformpressure intervals (referred to as "pressure weighting"in the table) or from values at uniform height intervalsthat were weighted by the density at those heights (referred to as "density weighting" ). The data compiled in Table A1 are summarized inTable 1. REFERENCESAchtemeier, G. L., 1969: Some observations of splitting thunder storms over Iowa on August 25-26, 1965. Preprints, Sixth Conf. on Severe Local Storms, Chicago, Amer. Meteor. Soc., 89-94.Bluestein, H. B., and C. J. Sohl, 1979: Some observations of a split ting severe thunderstorm. Mon. Wea. Rev., 107, 861-878.--, and G. R. Woodall, 1990: Doppler-radar analysis of a low precipitation severe storm. Mon. Wea. Rev., 118, 1640-1664.Boe, B. A., and H. L. Johnson, 1990: Destabilization antecedent to a tomadic northern High Plains mesoscale convective system: A case study. Preprints, 16th Conf. on Severe Local Storms, Kan anaskis Park, Alberta, Canada, Amer. Meteor. Soc., 538-541. , J. L. Stith, P. L. Smith, J. H. Hirsch, J. H. Helsdon, Jr., A. G. Detwiler, H. D. Orville, B. E. Martner, R. F. Reinking, R. J. Meitin, and R. A. Brown, 1992: The North Dakota Thunder storm Project: A cooperative study of High Plains thunder storms. Bull. Amer. Meteor. Soc., 73, 145-160.Brown- R. A., 1992: Initiation and evolution of updraft rotationwithin an incipient supercell thunderstorm. J. Atmos. Sc-, 49,1997-2014. , 1993: A compositing approach for preserving significant fea tures in atmospheric profiles. Mon. Wea. Rev., 121, 874-880. , and V. T. Wood, 1983: Improved severe storm warnings using Doppler radar. Natl. Wea. Dig., 8(3), 17-27; Errata, 9(1), 2. , and , 1991: On the interpretation of single-Doppler ve locity patterns within severe thunderstorms. Wea. Forecasting, 6, 32-48. - , D. W. Burgess, and K. C. Crawford, 1973: Twin tornado cy clones within a severe thunderstorm: Single Doppler radar ob servations. Weatherwise, 26, 63-~69, 71. , C. R. Safford, S. P. Nelson, D. W. Burgess, W. C. Bumgamer, M. L. Weible, and L. C. Fortner, 1981: Multiple Doppler radar analysis of severe thunderstorms: Designing a general analysis system. NOAA Tech. Memo. ERL NSSL-92, National Severe Storms Laboratory, Norman, OK, 18 pp. [NTIS PB82-113117.]Burgess, D. W., 1981: Evidence for anticyclonic rotation in left-mov ing thunderstorms. Preprints, 20th Conf. on Radar Meteorology, Boston, Amer. Meteor. Soc., 52-54. , and E. B. Curran, 1985: The relationship of storm type to environment in Oklahoma on 26 April 1984. Preprints, 14th Conf. on Severe Local Storms, Indianapolis, IN, Amer. Meteor. Soc., 208-211. , L. R. Lemon, and G. L. Achtemeier, 1976: Severe storm split ting and left-moving storm structure. The Union City Tornado of 24 May 1973, NOAA Tech. Memo. ERL NSSL-80, National Severe Storms Laboratory, Norman, OK, 53-66. [NTIS PB 269443/AS.]Charba, J., and Y. Sasaki, 1971: Structure and movement of the se vere thunderstorms of 3 April 1964 as revealed from radar and surface mesonetwork data analysis. J. Meteor. Soc. Japan, 49,Conway, J. W., and M. L. Weisman, 1988: An investigation into the splitting and propagation of the 13 June 1984 Denver hailstorms. Preprints, 15th Conf on Severe Local Storms, Baltimore, MD, Amer. Meteor. Soc., 276-279.Davies-Jones, R. P., D. Burgess, and M. Foster, 1990: Test of helicity as a tornado forecast parameter. Preprints, 16th Conf. on Severe Local Storms, Kananaskis Park, Alberta, CanadacAmer. Meteor. Soc., 588-592.Donaldson, R. J., Jr., 1970: Vortex signature recognition by Doppler radar. J. Appl. Meteor., 9, 661-670.Eilts, M. D., and S. D. Smith, 1990: Efficient dealiasing of Doppler velocities using local environment constraints. J. Atmos. Oce anic Technol., 7, 118-128.Fujita, T., and H. Grandoso, 1968: Split of a thunderstorm into anticyclonic and cyclonic storms and their motion as determinedfrom numerical model experiments. J. Atmos. Sci., 25, 416439.Haglund, G. T., 1969: A study of a severe local storm of 16 April 1967. ESSA Tech. Memo. ERLTM-NSSL 44, National Severe Storms Laboratory, Norman, OK- 54 pp. [ NTIS PB-184970.]Hammond, G. R., 1967: Study of a left moving thunderstorm of 23 April 1964. ESSA Tech. Memo. IERTM-NSSL 31, National Severe Storms Laboratory, Norman, OK, 75 pp. [NTIS PB 174681 .]Harrold, T. W., 1966: A note on the development and movement of storms over Oklahoma on May 27, 1965. ESSA Tech. Memo. IERTM-NSSL-29, National Severe Storms Laboratory, Nor man, OK, 1-8. [NTIS AD-644899.]Hitschfeld, W., 1960: The motion and erosion of convective storms in severe vertical wind shear. J. Meteor., 17, 270-282.Jorgensen, D. P., P. H. Hildebrand, and C. L. Frush, 1983: Feasibility test of an airborne pulse-Doppler meteorological radar. J. Cli mate Appl. Meteor., 22, 744-757.Kingsmill, D. E., and R. M. Wakimoto, 1991: Kinematic, dynamic, and thermodynamic analysis of a weakly sheared severe thun derstorm over northern Alabama. Mon. Wea. Rev., 119, 262 297.Kubesh, R. J., D. J. Musil, R. D. Farley, and H. D. Orville, 1988: The 1 August 1981 CCOPE storm: Observations and modeling results. J. Appl. Meteor., 27, 216-243.Lemon, L. R., and D. W. Burgess, 1980: Magnitude and implications of high speed outflow at severe storm summits. Preprints, 19th Conf. on Radar Meteorology, Miami Beach, FL, Amer. Meteor. Soc., 364-368. --, and R. A. Brown, 1978: Tornadic storm airflow and morphology derived from single-Doppler radar measurements. Mon. Wea. Rev., 106, 48-61.Marwitz, J. D., 1972: The structure and motion of severe hailstorms. Part I: Supercell storms. J. Appl. Meteor., 11, 166-179.Newton, C. W., and J. C. Fankhauser, 1964: On the movements of convective storms, with emphasis on size discrimination in re lation to water-budget requirements. J. Appl. Meteor., 3, 651 668.Rotunno, R., and J. B. Klemp, 1982: The influence of the shear induced pressure gradient on thunderstorm motion. Mon. Wea.SEPTEMBER 1994 BROWN AND MEIT~N 2067Thorpe, A. J., and M. J. Miller, 1978: Numerical simulations showing the role of the downdraught in cumulonimbus motion and split ting. Quart. J. Roy. Meteor. Soc., 104, 873-893.Vasiloff, S. V., M. H. Jain, D. L. Keller, A. Witt, V. T. Wood, P. L. Spencer, G. J. Stumpf, and M. D. Eilts, 1993: An evaluation of two Doppler radar mesocyclone detection algorithms. Preprints, 26th Int. Conf on Radar Meteorology, Norman, OK, Amer. Me teor. Soc., 657-659.Weaver, J. F., 1979: Storm motion as related to boundary-layer con vergence. Mon. Wea. Rev., 107, 612-619.Wilhelmson, R. B., and J. B. Klemp, 1981: A three-dimensional nu merical simulation of splitting severe storms on 3 April 1964. J. Atmos. Sci., 38, 1581-1600.--, and C.-S. Chen, 1982: A simulation of the development of successive cells along a cold outflow boundary. J. Atmos. Sci., 39, 1466-1483.Wilk, K. E., 1966: Motion and intensity characteristics of the severe thunderstorms of April 3, 1964. ESSA Tech. Memo. IERTM NSSL-29, National Severe Storms Laboratory, Norman, OK, 9 21. [NTIS AD-644899.]