P7.5�� WSR-88D MESOCYCLONE CHARACTERISTICS OF
SELECTED THUNDERSTORMS
�DURING THE SOUTHWEST GEORGIA TORNADO OUTBREAK
ON 13-14 FEBRUARY 2000
T. J. Turnage*
Robert R. Lee
�����������������������������������������������������������
E. Dewayne Mitchell
NOAA / National Severe Storms Laboratory, Norman, Oklahoma
1.
INTRODUCTION
�On 13-14 February 2000, at least four strong tornadoes swept
across portions of southwest Georgia.�
The hardest hit community was Camilla, which was struck by a tornado
estimated to be of F3 intensity on the Fujita scale (see Fig. 1 in Gould et al.
2000 in this Preprint volume).� This
tornado produced 10 fatalities.�
Approximately 45 minutes later, another supercell thunderstorm produced
a tornado with an identical track to the first, and resulted in an additional
fatality in Camilla. Yet another supercell later produced two tornadoes 40
minutes apart in Grady County and along the Colquitt/Tift County line near
Omega (Fig. 1 in Gould et al. 2000).�
Detailed aerial surveys allowed the tornado tracks to be resolved
accurately.� These mesocyclones occurred
at ranges from the Tallahassee WSR-88D Doppler radar (KTLH WSR-88D) that varied
from 70 km for the Grady County tornado to 127 km for the Omega tornado.
This paper will investigate trends of depth, rotational
velocity, and shear of these tornadic mesocyclones as depicted by the KTLH
WSR-88D. Selected output parameters from the NSSL Mesocyclone Detection
Algorithm (MDA; Stumpf et al. 1998) will then be assessed to determine which
parameters, if any, could be effective predictors of tornadoes for these
events.� Finally, the WSR-88D Build 10
Tornado Detection Algorithm (TDA; Mitchell et al. 1998) will be scored using
procedures developed by the Operational Support Facility (OSF) and the National
Severe Storms Laboratory (NSSL), and an adaptable parameter set optimized for
this severe weather episode will be developed and discussed.
2. METHODOLOGY
KTLH WSR-88D Archive level II data were collected and
analyzed using WATADS version 10.2 (NSSL 2000). The pertinent mesocyclones were
identified in reflectivity and storm-relative velocity product animations.� The mesocyclone tracks were correlated by
location and by WSR-88D volume scan times with the specific damage tracks
assessed from aerial surveys.� Once
tornado occurrence times were determined in this way, time-height trends of
mesocyclone rotational velocity, gate-to-gate velocity, shear, and diameter
were plotted using the WATADS trend feature.�
Finally, the TDA was scored using procedures highlighted in the WSR-88D
Algorithm Scoring Procedures (OSF 1999).
3. RESULTS
The first mesocyclone analyzed was associated with the first
Camilla tornado.� The WSR-88D depicted a
steady increase in mesocyclone depth prior to the tornado, with a steady
decrease in depth during the tornado's lifetime (not shown).� Around 0421 (all times UTC), about 20
minutes before the Camilla tornado, there was a maximum of rotational
velocities > 35 m/s that appeared simultaneously in the lowest elevation
slices (Fig. 1).� Damage surveys were
not taken of the area the mesocyclone passed over at this time, so it is
possible that these higher rotational velocities were associated with an
unreported tornado.
Figure 1 also shows that slightly higher rotational
velocities appeared to descend from the 1.5 degree elevation slice at 0436 just
prior to touchdown of the Camilla tornado.�
However, improperly dealiased velocities were depicted by the WSR-88D at
this slice and time (not shown), so it is theorized that the descent was
falsely implied by this dealiasing problem.
This theory is corroborated in Figure 2, which indicates an
abrupt and simultaneous increase of gate-to-gate velocities in the lowest two
elevation slices. Shear trends (Fig. 3) show values peaking in the lowest
elevation slice at the onset of the Camilla tornado,� although in the earlier strong circulation occurring around 0421,
a descent of higher shear values is evident.
Next, the second tornadic Camilla mesocyclone was
investigated.� Aerial and ground surveys
were unable to resolve an individual damage track with this tornado because it
overlapped the track of the first. Starting and ending times for this second Camilla
tornado therefore remain uncertain; however, rotational and gate-to-gate
velocity trends (Figs. 4 and 5) strongly suggest a tornado was on the ground by
0525, about 15 minutes prior to the final fatality in Camilla.
Figure 1.� Time series plot of WSR-88D detected rotational velocity in m /
s.
Mesocyclone depth rapidly fluctuated prior to and during the
probable time of this tornado (not shown), and Figures 4 and 5 show that the
detected mesocyclone base dropped dramatically around 0510. These figures also
show stronger and deeper rotation compared to the first Camilla tornado.� With a track virtually identical to the
first tornado, the second Camilla tornado should have been sampled by the
WSR-88D equally well.� As in the case of
the first tornado, rotational and gate-to-gate velocities increased
simultaneously in the lowest elevation slices, most notably in the lowest 4
slices of Figure 5. Unlike the first Camilla tornado, the highest shear and
lowest diameter values remained aloft in the second and third elevation slices
while lowest diameters were correspondingly greater (Figs. 6 and 7).
Figure
2.� Same as Figure 1, but for
gate-to-gate velocity.
Figure
3.� Same as Figure 2, but for shear.
Next, the mesocyclone that produced a tornado in northern
Grady County was analyzed.� Figure 8
shows higher rotational velocities at the 1.5 degree elevation slice lasting
for three volume scans prior to tornado touchdown.� High gate-to-gate velocities of at least 50 m/s similarly spread
down from the 1.5 degree slice (not shown).�
Figures 9 and 10 show a descent of high shear and small diameter
immediately before the tornado.� In
fact, the mesocyclone diameter at the 0.5 degree elevation slice shrank from 8
km to 3 km just prior to the tornado.�
This descent of tightening circulation was pronounced enough to cause a
corresponding (increase, decrease) in MDA derived (depth, base height) just
prior to the tornado (Fig. 11).
Figure
4.� Same as Fig. 1, but for the second
Camilla tornado.� The estimated time of
the additional damage and fatality in Camilla is labeled with a bar.�
Figure 5.� Same as Fig. 4, but for gate-to-gate velocity.
The supercell that produced the Grady County tornado
produced another tornado near Omega from 0642 to 0648.� The last 0.5 degree slice of the associated
mesocyclone was detected by the WSR-88D MDA at 0633.� At 0645, the MDA assigned a new mesocyclone identifier to the
circulation responsible for the Grady County tornado, despite the fact velocity
data showed a persistent rotational couplet this whole time.� The MDA�s problem with mesocyclone
continuity greFFCy hindered the effectiveness of the trend feature for this
tornado.� This effect was noted with
other mesocyclones for this event when increasing range or multiple mesocyclone
detections played a role.
Finally, a Tornado Detection Algorithm (TDA) adaptable
parameter set (APS) was optimized for this tornadic outbreak.� Each APS consists of a combination of
thresholds for the Tornado Vortex Signature (TVS) depth, Low-Level
Delta-Velocity (LLDV), and Maximum Delta-Velocity (MXDV) detected within the
TVS.
Table 1 compares the four operationally approved APSs with
the one optimized for this event.� The
high event thresholds (bottom row) emphasize the large and strong nature of the
circulations that occurred. In fact, thresholds for this outbreak even exceed
those for the Isolated Supercell set (labeled Super in Table 1).�
The Default and Squall Line/Tropical APSs (Def and Squall,
respectively, in Table 1) both had an identically high Probability of Detection
(POD) for this event, but because thresholds were too low, each False Alarm
Rate (FAR) was higher, resulting in a lower Critical Skill Index (CSI) and
Heidke Skill Score (HSS) compared to the optimized APS.� It is interesting to note how similarly
these two APSs performed.� Even for this
strong event, the Minimized set (Min in Table 1) had thresholds set too high,
as evidenced by a very low POD, CSI, and HSS.
Figure 6.� Same as Fig. 4, but for shear.
4. CONCLUSIONS
WSR-88D algorithm-derived trends of depth, shear, and
rotational velocity were investigated for three mesocyclones.� The best predictor for tornado occurrence
was rotational velocity exceeding 35 m s-1 in the lowest elevation
slice.� This predictor was corroborated
by the TDA APS optimized for this event.�
For both Camilla tornadoes, rotational velocities increased abruptly and
simultaneously in the lowest elevation slices, which would have resulted in late
warnings had this predictor been used alone.
Figure 7.� Same as Fig. 4, but for diameter.
The rapid development of the Camilla tornadoes in the lowest
elevation slices suggests a boundary layer mechanism strongly contributed to
tornadogenesis. The Grady County tornado showed a more classic descent of
higher rotational velocities and a well-defined increase of shear in the lowest
elevation slice corresponding to a significant decrease in mesocyclone
diameter.� Surface mesoanalyses of
southwest Georgia during the event revealed an outflow boundary the
mesocyclones crossed prior to tornadogenesis (not shown; Fournier 2000).� Such boundary interactions leading to
tornadogenesis have been documented in other studies (e.g., Maddox 1980,
Rasmussen 2000).
Figure 8.� Same as Fig. 1, but for the Grady County
tornado.
Mesocyclone depth trends varied prior to tornadogenesis for
each case, so depth alone was not a good predictor of tornadoes; however a
gradual decrease in depth was noted in all cases while the tornadoes occurred.
The APS optimized for this event suggests the tornadic
circulations that occurred were qualitatively comparable to those found with
isolated supercells of the Great Plains.�
Use of this optimized APS would likely result in a low POD for the types
of tornadoes that normally occur in the Tallahassee County Warning Area.� Future work will focus on developing an APS
that represents a broader spectrum of local cases.
Figure 9.� Same as Fig. 6, but for the Grady County tornado.
Overall, the WATADS trend feature for this outbreak was very
useful, but mesocyclone continuity problems were noticed with mesocyclones
poorly defined by increased range or multiple detections.� Real-time use of this feature in the Warning
Decision Support System (NSSL 2000) would likely help severe weather operations
significantly.
Acknowledgments:� A special thanks to Irv Watson (NWS
Tallahassee SOO) for his assistance with the figures and helpful suggestions.
5.
REFERENCES
�Available on
request.
Figure 10.� Same as Fig. 7, but for the Grady County tornado.
Figure 11.� Mesocyclone base height and depth for the Grady County tornado.
Parameters |
Scores |
||||||
APS |
Depth (km) |
LLDV (m/s) |
MXDV (m/s) |
POD |
FAR |
CSI |
HSS |
Min |
5.0 |
56 |
74 |
.13 |
.25 |
.13 |
.18 |
Def |
1.5 |
25 |
36 |
.91 |
.56 |
.42 |
.41 |
Squall |
1.6 |
27 |
27 |
.91 |
.58 |
.40 |
.39 |
Super |
3.1 |
27 |
30 |
.86 |
.54 |
.43 |
.43 |
Event |
3.2 |
39 |
43 |
.78 |
.31 |
.58 |
.65 |
Table
1.� WSR-88D TDA Adaptable Parameter Set
Comparison.
�* Corresponding author address:� T. J. Turnage, National Weather Service, 3300 Capital Cir. SW,
Tallahassee, FL 32310;� e-mail:
Thomas.Turnage@noaa.gov