P4.7 OBSERVATIONS OF THE 17 JUNE 1997 TORNADOES
Patrick J. Spoden*, Timothy W. Troutman**, S. Douglas Boyette*, David L. Humphrey*,
Paul G. Witsaman*, Jason B. Wright**
*NWSO Paducah, Kentucky **NWSO Nashville, Tennessee
Blluestein (1985) introduced the term "landspouts" to describe a tornado developing under a line of rapidly growing cumulus towers owing its similarity to the formation of waterspouts. The formation of landspouts also was hypothesized by Bluestein to be related to local stretching of vorticity by intersecting outflows. Brady and Szoke (1989) further presented evidence of "a class of weaker non-mesocyclone tornadoes which 1) are forced by mesoscale boundary layer interactions and processes, and 2) generally develop in benign (weakly sheared and weakly forced) synoptic environments.
Doswell et al., (1990) presented three conceptual models for supercell thunderstorms including classical, low-precipitation, and high precipitation. Davies (1993) produced a detailed description of a miniature supercell storm (hereafter MS) in the Central Plains and Foster et al., (1995) documented their structure and evolution in northern Texas. Numerical simulations of MS storms have been made by Stalker et al. (1993) and Wicker and Cantrell (1996). Burgess et al., (1995) studied MS storms as seen by WSR-88D radars across different parts of the country while Grant & Prentice (1996) have produced statistics dealing with MS mesocyclones. Even with these helpful studies, MS storms present the additional challenge of operator recognition beyond medium ranges. Burgess et al (1995) noted that detections of MS storms may be difficult at ranges beyond 167 km (90 nm) and a loss of resolution is possible even beyond 102 km (55 nm).
On the evening of 17 June 1997, a series of six tornadoes touched down in the Lower Ohio and Tennessee Valleys (Fig. 1). All were of F1 or F0 strength on the Fujita scale and none were on the ground for more than two minutes, however, damage estimates were near one million dollars. Tornadoes were not anticipated on this day by the operational forecasters. This study will focus on the environment in which these storms formed, how the numerical models handled the forecast of the environment, and a radar perspective of 3 of the 4 storms analyzed (two of the storms produced multiple tornadoes). The purpose of this study is to attempt to find ways to forecast these types of storms better and to aid in the recognition of MS storms on the WSR-88D.
Figure 1. Map of tornado reports
At 1200 UTC, 17 June 1997, the lower Ohio and Tennessee Valleys were under the influence of an area of low pressure, located to the west. A positively tilted 500 mb trough extended from southeast lower Michigan through central Illinois, to a weak 500 mb low circulation centered over northeast Oklahoma.
A 26 m s-1 speed maximum was observed on the 500 mb analysis at 1200 UTC in the vicinity of Lake Charles, Louisiana. The 1200 UTC 850 mb analysis of the region indicated that a moisture axis extended northeast from central Mississippi into middle Tennessee. The nose of a 23 m s-1 850 mb jet was advancing northeast from Shreveport, LA (SHV) into Mississippi and Alabama. A dry slot analyzed from 300 to 850 mb was also being advected northeast out of central Texas into the Lower Ohio Valley region. At the surface, one weak trough extended from central Kentucky to southwest Arkansas, where an area of low pressure was analyzed. Another weak trough extended east from this low through northern Mississippi and Alabama.
Satellite imagery loops revealed the presence of a mid-level dry slot rotating around a circulation centered over south central Missouri. The dry slot rotated through Arkansas during the early afternoon hours, then progressed into southwest Kentucky. At 0000 UTC 18 June, a surface trough extended east from near Paducah to about 20 miles north of Bowling Green Kentucky. A 23-26 m s-1 speed maximum between 2.4 and 3.0 km located over the region was shown by the 0000 UTC Nashville, TN (BNA) sounding (not shown). Table 1 lists various convective parameters from both the BNA and LIT soundings.
3. NUMERICAL MODEL COMPARISON
This section examined the model output that would have been available to
the operational forecasters during the mid-afternoon hours on 17 June, 1997. This included
the 1200 UTC NGM, Early-
Eta, and the 1800 UTC RUC. Fields studied included 500 mb heights, vorticity, and temperatures, 500 and 250 mb isotachs, 850 mb isotachs and moisture fields, 1000 to 500 mb mean relative humidity, and 0-3 km storm relative helicity fields. Storm relative inflow at 850 mb and SR flow at 500 mb were examined using a PC-GRIDDS macro (Thompson, 1996). Bulk Richardson Shear, which considers shear over the lowest 6 km, was also examined via a PC-GRIDDS macro. Table 2 lists convective parameters from the models. The 12 hour forecast data were compared to satellite and 0000 UTC observations.
1200 UTC Eta model
A positively tilted 500 mb trough was forecast to lift from southern Oklahoma into northern Arkansas by 0000 UTC. PVA was forecast to be channeled. An 850 mb moisture axis was forecast from southern Arkansas into west Tennessee with dewpoints near 14o C. Mean relative humidity panels also depicted a dry slot rotating into extreme southwest Kentucky.
1200 UTC NGM model
As with the ETA model, a positively tilted 500 mb trough was forecast to lift from southern Oklahoma into northern Arkansas by 0000 UTC. PVA was forecast to be channeled. The 850 mb jet was forecast to be over southern Tennessee, farther south and weaker than the Eta. The 850 mb moisture axis was from southern Arkansas into southeast Tennessee with dewpoints near 16o C. Mean relative humidity panels also depicted the dry slot extending from southeast Missouri into central Kentucky by the time of the severe weather. This model had the best depiction of the dry slot of the three studied.
1800 UTC RUC model
A positively tilted 500 mb trough was forecast to be located over central Arkansas at 0000 UTC, a bit further south than the earlier Eta and NGM positions. Weak PVA was forecast over the western part of the region. The 850 mb jet was forecast to be centered over northern Alabama into middle Tennessee. The mean relative humidity panel showed the dry slot along the Gulf Coast.
|SR Helicity (m2 s2)||0-40||40-80||30-40|
|Favorable SR flow - 850 mb||N||N||N|
|Favorable SR flow - 500 mb||Y||Y||Y|
|850 Jet (m s-1)||10-13||15-18||10-15|
|500 Jet (m s-1)||23||23||26|
|250 Jet (m s-1)||33||41||36|
|Dry Slot near area||Y||Y||N|
4. RADAR OBSERVATIONS
WSR-88D Archive II data, utilizing WATADS 9.0 (1997), was used in this study due to incomplete archive IV data from the offices involved. Data from KHPX (Ft. Campbell) was used for the Central Christian County and Oak Grove storms, data from KPAH was used for the Evansville storm. When analyzing archive II data, a few important features were much clearer than when utilizing archive IV data. This will be pointed out, where appropriate, in the following discussions.
Storm #1 - Central Christian County Storm
This storm developed ahead of a short squall line in north central Tennessee at 2115 UTC (all times hereafter UTC). The cell was vertically stacked with weak rotation of 10 m s-1 noted at a height of 1.6 km. By the 2157 volume scan a WER (weak echo region) was detected along the southeast flank of the cell with a base of about 2.1 km. Maximum rotational velocities (Vr) were 11 m s-1 and extended from 1.1 km to 3.0 km in depth. Between 2157 and 2215 a BWER (bounded weak echo region) developed along the southeast flank, the circulation depth increased to 3.1 km with maximum Vr values of 16 m s-1. The height of the 50 to 55 dBZ reflectivity core was 4.1 km, its greatest extent so far. Figure 2 is a cross section of the storm at 2215.
By 2221, the storm had weakened and there were no indications of a BWER. This trend continued through 2250. During this time, maximum Vr values were generally in the 5 to 8 m s-1 range. Rapid re-organization occurred between 2250 and 2256. The reflectivity data revealed a kidney bean shape (Molleret al., 1990) with a WER developing on the southeast flank of the storm. The circulation had tightened during the 2256 volume scan. By 2302, a large WER was present, capped at 4.8 km, and the cell had become more supercellular in appearance with a hook evident in the reflectivity data. Velocity data indicated convergence on the lowest elevation angle, but deep rotation extended through 4.8 km with a maximum Vr of 8 m s-1. Maximum Vr values then increased from 13 m s-1 during the 2308 scan, to 19 m s-1 during the 2320 scan, to a maximum of 22 m s-1 by 2326. An F1 tornado touched down in central Christian County Kentucky around 2320. By 2328, the maximum Vr decreased to 11 m s-1, the hook had become ill-defined, and the reflectivity pattern was elongated. The WER diminished by 2338. Between the 2344 and the 0043 volume scan the storm went through variations of intensity. The reflectivity data quickly organized again with a WER present (capped at 2.1 km) during the 0049 to 0101 volume scans. During this time, maximum Vr values varied from 13 to 16 m s-1. An F0 tornado briefly touched down from this cell between the 0101 and 0107 volume scans. By 0107, the WER had dissipated, however, the Vr values remained essentially unchanged. This storm continued to move northeast but did not produce any additional severe weather.
Figure 2 Cross section of central Christian county
Storm. Y-axis in k-ft, X-axis in nm.
Storm #2 - Oak Grove Storm
The Oak Grove storm produced two tornadoes along the Kentucky/Tennessee border between 0013 and 0023. The cell which produced the tornadoes developed between 2344 and 2350. The origins of the circulation that was traced to these tornadoes, however, developed long before the actual cell. Rotational velocities with this circulation never extended beyond 4 km in depth. Figure 3 details the track of the circulation along with cell descriptions.
At 2224, a 34-39 dBZ echo developed within a group of disorganized cells in central Stewart County in northern Tennessee. Rotation developed with this cell at 2244 with maximum Vr values of 6 m s-1 at a height of 2.7 km. By 2256, the cell was strongly tilted to the northeast with an elevated 50 dBZ core. The storm continued to increase through 2308 as maximum Vr values increased to 10 m s-1. By 2326, the core had weakened and there were two separate 40 to 45 dBZ reflectivity cores, which eventually decreased to a cluster of weak cells. The rotation, however, continued with a maximum Vr of 8 m s-1 through a depth of 1.1 km, about 6.5 km south of the highest reflectivities.
At 2344, along the southwest flank of the cluster, a 34 to 49 dBZ core developed in conjunction with the circulation, which had increased to 14 m s-1 through a depth of 2.6 km. Weak reflectivities along the southwest flank indicated the position of a rear flank downdraft. Rapid development continued into the 2350 volume scan. A 50 to 60 dBZ core extended from 1.2 km to 2.2 km with a rear flank downdraft still evident. At 2356, a BWER developed along the southern flank with a cap between 1.5 km and 2.1 km and a diameter of 3.7 km. The diameter of the entire storm was about 13 km. During the 0001 and 0007 volume scans, the BWER remained the same size, with a rear flank downdraft evident. Maximum Vr declined slightly to 10 m s-1 but the depth continued to be about 2.5 km. No clear rotation was detected on the lowest elevation angle. At 0013, the storm began to dissipate with the BWER collapsing to only 1.2 km. Maximum Vr continued at 10 m s-1 through a depth of about 2.5 km. No clear rotation was detected on the lowest 2 slices even though the storm was 13 km away from the KHPX WSR-88D. An F0 tornado touched down from 0013 UTC to 0015 UTC at Fort Campbell along the Kentucky/Tennessee line. The circulation trend continued into the 0019 UTC volume scan. The storm was in close proximity to the KHPX radar and the outline of the BWER was difficult to discern, although a rear flank downdraft could be seen. An F1 tornado touched down in Oak Grove, Kentucky for less than one minute at 0023. The cell continued northeast over the KPHX radar and continued to weaken, finally dissipating by 0049. The archive IV data (lowest four reflectivity slices) analyzed from this event did not always clearly reveal the presence of the BWER, but at times, a pendant was indicated.
Storm #3 - Evansville Storm
The Evansville storm was much farther away from a WSR-88D than the previous storms discussed. Because of this distance (148 to 166 km), low-level details detected in the previous storms were not seen in the Evansville storm. However, some useful details were extracted from the data.
The initial storm developed around 2200 in eastern Henderson County Kentucky with a vertically stacked 50 dBZ reflectivity core. This was about 50 km south of a band of showers and thunderstorms located over southern Illinois and southern Indiana. Weak Vr values of 5 m s-1 were detected at a height of 3.7 km in the intial cell. By 2205, the high reflectivity core (55 dBZ) increased vertically and extended from 1.8 km to 3.7 km. The top of the storm was about 5.8 km. Through 2217, little change was noted in the storm, which remained vertically stacked with little or no rotation. At 2223, the storm split into a northern (cell #1) and southern (cell #2) storm. No rotation was detected, the cells remained vertically stacked, and the top was about 6.1 km. These cells crossed the Ohio River into southwest Vanderburgh County at 2229, and although individual cores were seen at higher elevation angles, the cells were joined together at the lowest slice. At 2235 another cell (#3) began to develop 11.1 km to the east of cells #1 and #2.
By 2246, cell #3 had increased in intensity to 45 dBZ. No rotation was detected in any of the cells, with tops around 6.7 km. Data from 2258 indicated that cell #3 had become the strongest cell with a 50 dBZ core up to 4.4 km and weak Vr values of 6 m s-1 detected at 2.2 km and 4.4 km. No rotation was detected in cells #1 or #2. At 2304, cells #2 and #3 began to merge as cell #1 moved to the north and weakened. The only rotation detected was in cell #3 at 2.4 km with a Vr of 3 m s-1. By 2310, weak rotation of 3 m s-1 was detected at 2.4 km, 4.8 km, and 7.3 km, however, it should be noted that this was near range obscured and incorrectly dealiased velocity data.
At 2316 the lowest slice at 2.4 km revealed two cores joined by 50 dBZ echo, with distinct 45 dBZ cores seen at 4.9 km. Tops of the cells were 9.5 km as Vr values increased slightly to 5 m s-1 at both 2.4 km and 4.9 km. The merger appeared to be complete by the 2321 volume scan. Maximum Vr values of 5 m s-1 remained at both 2.4 km and 4.9 km and were located in the southeast portion of the cell (formerly cell #3). At 2327, one minute before tornado touchdown, the cells appeared to split, with the northernmost cell the strongest in intensity. Maximum Vr in the southernmost storm remained at 5 m s-1 at a height of 2.4 km, but increased to 8 m s-1 at a height of 4.9 km. By 2333, maximum Vr values increased to 8 m s-1 at a height of 2.7 km with a weak divergent signature at 5.4 km in the southern cell. By 2339, the northernmost cell continued to move north into southern Gibson County. Weak Vr values of 5 m s-1 at a height of 2.8 km, and 3 m s-1 at 5.6 km continued to be detected with the slower, southernmost cell. Although the radar data did not reveal any supercell characteristics, photographs (not shown) of the tornado revealed a wall cloud with a dry slot wrapping into the storm.
Forecasters were caught unaware that rotating thunderstorms could develop in this type of environment. Using numerical model data, some of the traditional tools used for the forecasting of rotating thunderstorms (i.e., storm-relative helicity and storm-relative inflow) did not clearly suggest that tornadoes were likely to occur in the lower Ohio Valley. However, the use of BRN shear did (in hindsight) point to the possibility of rotating thunderstorms. Forecasters should be aware of this possiblility when an upper level circulation is nearby along, and unusually cold temperatures are present aloft.
Even if the possibility of rotating thunderstorms is established, the problem of identifying MS storms using the WSR-88D network remains. It is clear from this case that all radars in and around the county warning area must be monitored for the development of rotating thunderstorms. Radar operators should look for reflectivity characteristics similar to classic supercells, but reduced in size. Only one storm in this study reached mesocyclone criteria according to the nomogram by the OSF (Andra et al. 1994). False alarms would be extremely high if warnings were issued each time forecasters used the WSR-88D data alone for the Oak Grove and Evansville storm without good spotter reports.
On this day, there were reports of funnel clouds earlier in the day, approximately two to three hours before these storms even developed. No reports of funnel clouds or tornadoes from these three storms were reported to NWSO PAH until the events were over.
Storm spotter reports must be taken seriously, even when no real identification of "tornadic" thunderstorms can be seen by the WSR-88D. Forecasters must be aware of the possibility of tornadoes in traditional "non tornado" environments and without significant radar signatures. In some cases, without a funnel cloud report before tornado touchdown, some MS tornadoes will be "unwarned", due to the short lifespan of the tornado. This will be the case until technology and research allows for a more definitive identification of MS storm characteristics.
The authors would like to thank Rich Thompson of SPC, David Andra, SOO NWSFO OUN, and Ron Przybylinski, SOO NWSFO STL, for their help in radar interpretation. We would also like to thank David Andra for his thorough review of this manuscript. MIC's Beverly A. Poole and Derrel R. Martin, along with the staffs of NWSO PAH and NWSO BNA need to be thanked for their unwavering support.
Available upon request.
Andra, D., V. Preston, E. Quoetone, D. Sharp, and P. Spoden, 1994: An operational
guide to configuring a WSR-88D principal user processor, Operations Training
Branch, Operational Support Facility
Bluestein, H.B., 1985: The formation of a "landspout" in a "broken-line" squall line in
Oklahoma. Preprints, 14th Conf. on Severe Local Storms, Indianapolis, IN.,
Amer. Meteor. Soc., 267-270.
Brady, R. H., and E. J. Szoke, 1989: A case study of nonmesocyclone tornado development in
northeast colorado: similarities to waterspout formation. Mon. Wea. Rev., 117, 843-856.
Burgess, D.W., and R.J. Donaldson, 1979: Contrasting tornadic storm types. Preprints,
11th Conf. on Severe Local Storms, Kansas City, MO, Amer. Meteor. Soc., 189-
_____, R. L. Lee, S. S. Parker, S. J. Keighton, and D. L. Floyd, 1995: A study
of mini supercells observed by WSR-88D radars. Preprints, 27th Conference
on Radar Meteorology, Vail, CO., Amer. Meteor. Soc.,
Davies, J. M., 1993: Small tornadic supercells in the central plains. Preprints, 17th
Conf. Severe Local Storms, St. Louis, MO., Amer. Meteor. Soc., 305-309.
Doswell III, C. A., A. R. Moller, and R. Przybylinski, 1990: A unified set of conceptual
models for variations on the supercell theme. Preprints, 16th conf. Severe
Local Storms, Indianapolis, IN., Amer. Meteor. Soc. 40-45.
Foster, M. P., A. R. Moller, L. J. Wicker, and L. Cantrell, 1995: The rapid evolution of a
tornadic small supercell; observations and simulation. Preprints, 14th Conf. on
Weather Analysis and Forecasting, Dallas, TX., Amer. Meteor. Soc. 323-328
Grant, B. and R. Prentice, :Mesocylone characteristics of mini supercell thunderstorms.
Preprints, 15th Conf. On Weather Analysis and Forecasting, Norfolk, VA, Amer.
Meteor. Soc., 362-365.
Moller, A. R., C. A. Doswell III, and R. Przybylinski, 1990: High-precipitation supercells: a
conceptual model and documentation. Preprints, 16th Conf. on Severe Local
Storms, Kananaskis Park, Alberta, Canada, Amer. Meteor. Soc., 52-57.
Stalker, J. R., K. R. Knupp, and E. W. McCaul, Jr., 1993: A numerical and
observational study of an atypical "miniature" supercell storm. Preprints, 17th
Conf. Severe Local Storms, St. Louis, MO., Amer. Meteor. Soc. 191-195.
Thompson, R. L., 1996: Supercell tornado forecasts derived from ETA model storm-
relative winds. Preprints, 18th Conf. Severe Local Storms, San Francisco, CA.,
Amer. Meteor. Soc., 362-366.
WATADS (WSR-88D Algorithm Testing and Display System) 1997: Reference Guide
for Version 9.0 [Available from Storm Scale Applications Division, National
Severe Storms Laboratory, 1313 Halley Circle, Norman, OK 73069.]
Wicker, L. J. and L. Cantrell, 1996: The role of vertical buoyancy distributions in
miniature supercells. Preprints, 18th Conf. Severe Local Storms, San Francisco,
CA., Amer. Meteor. Soc., 225-229.