A tornado ripped the roof from this house under construction in the
Barkers Ridge subdivision near Bessemer City, North Carolina, at 
810 PM EST on January 13, 2006.

Click here to view an image gallery showing additional storm damage.

1.  Introduction

A tornado touched down in Gaston County, North Carolina, just east of 
Bessemer City (Fig. 1) at approximately 810 PM EST on 13 January 2006 
(0110 UTC 14 January), producing damage of F1 intensity on the Fujita 
Scale.  [All times are referred to in Universal Time Coordinated (UTC) 
in this document, which Eastern Standard Time plus five hours.]  The 
damage path was approximately one-half mile long and 100 yards wide, 
and stretched in a southwest to northeast direction.  The tornado 
occurred 62 miles east northeast of the Weather Surveillance Radar-88 
Doppler (WSR-88D) located at the National Weather Service (NWS) Weather 
Forecast Office (WFO) in Greer, South Carolina (KGSP, not shown) and 
21 miles west of the Terminal Doppler Weather Radar (TDWR) located 
north of the Charlotte - Douglas International Airport (TCLT, Fig. 1). 

Figure 1.  Map of Gaston County, North Carolina.  Bessemer City is in 
the west central portion of the county.  The approximate location of 
the tornado is shown by the yellow box. The TCLT radar is located 
approximately 3 miles north of Paw Creek in Mecklenburg County.  Map
created with Delorme Street Atlas USA 2006.  Click on image to enlarge.

Forecasters in the western Carolinas have for many years recognized 
environments characterized by strong wind shear, weak instability, and 
strong deep layer forcing to be favorable for development of non-supercell 
tornadoes (NST).  The Bessemer City tornado developed within a quasi-
linear convective system (QLCS) ahead of a cold front accompanying the 
passage of a strong upper level trough.  The QLCS was quite shallow, as 
radar echo tops associated with the strongest updrafts rarely exceeded 
25,000 feet.  Radar reflectivity and radial velocity characteristics of 
the Bessemer City storm were similar to those observed during other non-
supercell tornado events in the western Carolinas.  Specifically, the  
reflectivity data displayed a "broken-S" pattern within the QLCS, similar 
to what has been documented in prior studies of NSTs (McAvoy et al. 2000).  
However, what sets this event apart from observations of other Broken-S-
type NSTs is the availability of data from the Charlotte TDWR, which may 
shed new light on tornadogenesis in these events.

2.  Synoptic Pattern and Stability Characteristics

The 500 mb analysis from the Storm Prediction Center (SPC) at 1200 UTC 
on 13 January 2006 indicated a highly amplified trough west of the 
Mississippi River Valley, with diffluent flow downstream of the trough 
over much of the eastern United States.  The surface analysis at 1200 UTC 
from the Hydrometeorological Prediction Center (HPC) showed a strong cold 
front extending from low pressure over Lake Superior, through the mid-
Mississippi Valley, into the western Gulf of Mexico.  By 2100 UTC, the 
cold front had moved east and extended from the Ohio Valley, through the 
Great Tennessee Valley into the eastern Gulf of Mexico (Fig. 2).

Figure 2.  HPC surface pressure and fronts analysis at 2100 UTC 
13 January 2006.  Observations are indicated by traditional station 
model.  Click on image to enlarge.

The upper trough at 500 mb moved east by 0000 UTC 14 January, providing 
a persistent diffluent flow across much of the eastern United States, 
including the Carolinas (Fig. 3).  A regional surface data plot at 
0100 UTC 14 January (Fig. 4) shows the wind, temperature, and dew point 
fields about ten minutes prior to tornado occurrence.  The airmass over 
the Carolinas was very warm and moist for early evening in January. 
Temperatures were in the lower 60s and dewpoint temperatures were in 
the upper 50s. An axis of strong convergence was implied along the cold 
front, with south southeast winds of 10 to 15 knots ahead of the wind 
shift, and southwest winds of 10 to 20 knots behind the front.  Strong
moisture convergence was indicated by the Storm Prediction Center (SPC) 
mesoscale analysis at 0100 UTC (Fig. 5).

Figure 3.  SPC objective analysis of 500 mb geopotential height, 
temperature, and wind for 0000 UTC 14 January.  Click on image 
to enlarge.

Figure 4.  Regional surface plot with surface front analysis at 0100 UTC 
14 January.  Click on image to enlarge.

Figure 5.  SPC objective analysis of surface moisture convergence and 
mixing ratio at 0100 UTC 14 January.  Click on image to enlarge.

Despite the moist and relatively warm air mass, instability was limited 
due to a deep layer of relatively weak temperature lapse rate indicated 
by the observed Greensboro, North Carolina (GSO), upper air sounding at 
0000 UTC on 14 January (Fig. 6).  However, a dry air intrusion in the 
850 mb to 700 mb layer steepened lapse rates sufficiently for weak 
destabilization to occur, as the sounding yields 198 J kg-1 of surface-
based Convective Available Potential Energy (CAPE).  A North American 
Mesoscale (NAM) initial hour sounding analysis (0000 UTC) showed the 
stability and wind characteristics at Charlotte (CLT, Fig. 7), which is 
approximately 17 miles east southeast of the tornado location.

Figure 6.  Skew-T log P diagram of the upper air sounding for GSO at 
0000 UTC 14 January 2006  (left) and severe weather indices (right).  
Click on images to enlarge.

Figure 7.  Bufkit display of initial hour NAM analysis for CLT.  
Click on image to enlarge.

The most notable aspect of both soundings is a wind profile characterized 
by strong veering and a rapid increase in wind speed with height. The 
0-1 km storm relative helicity (SRH) at GSO is 283 m2s-2. This is well 
above the median value of 0-1 km SRH of 137 m2s-2 found to be associated 
with weak tornadoes based on a study of 916 soundings proximal to tornado 
occurrence (Thompson et al. 2003). 

The Day 1 Convective Outlook issued by the SPC at 2000 UTC highlighted 
the environment characterized by strong shear and weak instability to 
the east of the Appalachians and included most of the Carolinas in a 
slight risk of severe thunderstorms. 

3.  Convective Evolution

During the mid afternoon, a large area of stratiform precipitation with 
embedded convection moved across the western Carolinas in response to 
deep synoptic scale forcing associated with the mid and upper level 
trough (not shown.)  However, by late afternoon, a mid level dry slot 
began to overspread this activity (Fig. 8) resulting in a reduction in 
precipitation coverage across Upstate South Carolina.  Meanwhile, a line 
of convection began to intensify along the back edge of the deep forcing 
(coincident with the cold front) across northeast Georgia, possibly as a 
result of steeper mid-level lapse rates that developed due to the 
advection of drier mid-level air (not shown.)  The convection continued 
to intensify and organize as it moved across the upper Savannah River 
Valley into Upstate South Carolina during the early evening.

Figure 8.  HPC reanalysis of 500 mb relative humidity at 0000 UTC 
14 January.  Click on image to enlarge.

By 2327 UTC, the reflectivity image from KGSP (Fig. 9) indicated a QLCS 
extended across the middle of Upstate South Carolina.  At this time, 
reflectivity data from KGSP revealed a slight bulge in the convective 
line along the Greenville - Laurens county line.  Ten minutes later, the 
bulging segment had evolved into a break in the QLCS west of Woodruff 
near the Spartanburg - Laurens county border (Fig. 10).  Forecasters at 
WFO GSP and others have long attributed this "broken-S" signature to 
the occurrence of weak and occasionally strong tornadoes in the eastern 
United States (McAvoy et al. 2000, Grumm and Glazewski 2004.)  Velocity 
data at this time (Fig. 11) indicated a line of relatively strong 
convergence associated with the high reflectivity region of the QLCS, 
with an area of weak shear across the line-break.  This is not unusual. 
Previous research (McAvoy et al. 2000) has suggested that intense 
vortices on the mesoscale (i.e., mesocyclones) have preceded "broken-S" 
tornadoes only in very rare cases.

Click here to view a 33 frame java loop of KGSP 0.5 degree base reflectivity.

Figure 9.  Base reflectivity on 0.5 degree scan from KGSP radar at 
2327 UTC 13 January.  Click on image to enlarge.

Figure 10.  As in Fig. 9, but for 2337 UTC.  Click on image to enlarge.

Figure 11.  Radial velocity on 0.5 degree scan from KGSP at 2337 UTC.  
Click on image to enlarge.

By 2342 UTC, the line-break across southern Spartanburg County persisted 
(Fig. 12), but was becoming ill-defined.  Concurrently, a pronounced bulge 
was developing in the convective line across Laurens County, northwest 
of Waterloo.  The reflectivity data indicated a narrow channel of weak 
reflectivity impinging on the back edge of the bulge from the southwest, 
implying a jet of subsiding, rear-flank inflow.  Meanwhile, storm-relative 
velocity data at 0.5 degrees revealed a weak rotational signature along 
the axis of the bulging segment (Fig. 13). By 2347 UTC, the rotation 
associated with the bulge had compressed and strengthened to 0.02 s-1 
(Fig. 14).  By 2349 UTC, this segment had evolved into a well-defined 
"broken-S" signature across central Laurens County (Fig. 15).

Figure 12.  As in Fig. 9, except for 2342 UTC.  Click on image to enlarge.

Figure 13.  As in Fig. 11, except for 2342 UTC.  Click on image to enlarge.

Figure 14.  As in Fig. 11, except for 2347 UTC.  Click on image to enlarge.

Figure 15.  As in Fig. 9, except for 2352 UTC.  Click on image to enlarge.

Despite the distinct appearance of the signature in Laurens County, a 
tornado apparently did not occur, nor was damage reported with the 
earlier signature across Spartanburg County.  Previous analysis of the 
"broken-S" pattern has revealed that they are often cyclical in nature.  
As the QLCS continued to move across Upstate South Carolina, the storm-
relative, forward flank inflow into the southern segment appeared to be 
unimpeded by rain-cooled air, unlike the northern part of the line, which 
was trailing an expansive area of stratiform precipitation.  For that
reason, forecasters’ attention remained focused on the evolution of the 
Laurens County segment, as opposed to the Spartanburg County segment. 

By 0047 UTC, the QLCS had moved within range of the TCLT radar.  At this 
time, the 0.2 degree base reflectivity image from TCLT (Fig. 16) indicated 
another "broken-S" signature in progress across northeast Cherokee County. 
Radar loops indicated this was the portion of the QLCS that broke over 
Spartanburg County earlier in the episode.  By 0053 UTC (Fig. 17), this 
feature had dissipated, although a channel of weaker reflectivity (less 
than 50 dBz) remained near the North Carolina - South Carolina border 
between two segments of higher reflectivity.  By 0059 UTC (Fig. 18), the 
reflectivity between the northern and southern segment continued to weaken, 
while the southern segment had pushed out slightly ahead of the northern 
one (Fig. 18).  This was approximately 10 minutes prior to tornado 
occurrence near Bessemer City. The 0.2 degree radial velocity image from 
TCLT at this time (Fig. 19, storm relative velocity was not available) 
indicated an area of convergence oriented from northeast to southwest 
extending from a pendant at the southern tip of the northern line segment. 
The inbound velocities associated with the convergence to the rear of the 
linear segment are coincident with a channel of minimum reflectivity, 
suggesting a subsident component to the flow.

Figure 16.  Base reflectivity on 0.2 degree scan from TCLT radar at 
0047 UTC 14 January.  Click on image to enlarge.

Figure 17.  As in Fig. 16, except for 0053 UTC.  Click on image to 
enlarge.

Figure 18.  As in Fig. 16, except for 0059 UTC.

Figure 19.  Radial velocity on 0.2 degree scan from TCLT radar at 
0059 UTC.

By 0105 UTC, a "broken-S" signature was evident across western Gaston 
County (Fig. 20) as the southern segment continued to move northeast 
slightly faster than the northern segment. However, there is a noticeable 
difference between the reflectivity pattern in Fig. 20 and those in 
Figs. 10 and 15. The "broken-S" patterns in Figs. 10 and 15 are of the 
"distinct" variety that have been documented in the scientific literature. 
In these cases, the high reflectivity (greater than 50 dBz) associated 
with the southern line segment extended north and east of the southern 
tip of the northern segment. This structure is not evident in Fig 20. 

There was a hint of cyclonic curvature in the reflectivity field at the 
southern tip of the northern segment. In addition, reflectivity continued 
to decrease within the channel between the 2 segments. Radial velocity 
(Fig. 21) indicated an area of low level convergence associated with an 
apparent descending rear inflow jet extending southwest from the northern 
segment. 

Figure 20.  As in Fig. 16, except at 0105 UTC.

Figure 21.  As in Fig. 19, except at 0105 UTC.

The 0111 UTC volume scan from the TCLT radar (Fig. 22) was the closest 
in time to the approximate time of tornado occurrence (likely 0109 to 
0110 UTC).  The reflectivity data depicted a narrow pendant at the 
southern tip of the northern segment.  A small indentation in the forward 
flank of the line segment was observed just north of the pendant.  This 
was possibly indicative of an area of easterly storm-relative inflow.  
It is interesting to note that the reflectivity east of the pendant had 
decreased to less than 20 dBz at this time.  This suggests that the 
descending rear inflow jet evident in the previous reflectivity images 
had turned cyclonically around the southern tip of the line segment.  
The radial velocity data continued to indicate a narrow band of outbound 
velocities extending southwest from the pendant.  Coincident with the 
pendant, the velocity data revealed a weak rotational signature (Fig. 23). 
However, the data was difficult to interpret due to the absence of a 
storm relative velocity product.  Subsequent images from TCLT actually 
depicted a hook-like appendage extending from the northern line segment 
(Fig. 24), although the significance of this is questionable as this 
feature became evident several minutes after the tornado had dissipated.

Figure 22.  As in Fig. 16, except at 0111 UTC.  The letter "T" indicates 
the approximate location of the tornado.

Figure 23.  As in Fig. 19, except at 0111 UTC.

Figure 24.  As in Fig. 16, except at 0112 UTC.

A series of images from KGSP prior to and during tornado occurrence 
(Figs. 25 and 26) reveal some significant differences with TCLT in 
regard to the detail of the structure of the evolving tornadic storm.  
Reflectivity images from KGSP did not reveal the cyclonic curvature in 
the southern tip of the northern segment just prior to tornadogenesis. 
In addition, the forward flank inflow notch and the hook-like structure 
that was evident in the TCLT images from 0111 to 0112 UTC was absent in 
the KGSP reflectivity field.  Although part of this may be attributed 
to the fact that the KGSP radar is much farther away from the storm 
than the TCLT radar (61 vs. 22 miles), it can also be said that the 
higher resolution provided by the 5 cm TDWR allows forecasters to see 
more detailed structure in precipitation systems than the WSR-88D.

Figure 25.  As in Fig. 9, but for 0107 UTC 14 January.  Click on image 
to enlarge.

Figure 26.  As in Fig. 11, but for 0112 UTC.  Click on image to enlarge.

The Bessemer City tornado was somewhat unusual in that it apparently 
occurred several volume scans after the line-break first became evident. 
Previous research has shown that tornadogenesis occurs as the line-break 
is developing, or shortly thereafter, providing at best 5 minutes of 
lead-time.  In this case, a warning issued at the time that the break 
first became evident in the reflectivity data would have provided 
approximately 10 minutes of lead time. However, this would have required 
real-time recognition of the evolving "broken-S" pattern, which was 
difficult in this particular case. 

4.  Summary 

The Bessemer City tornado of 13 January 2006 shared some of the radar 
characteristics of previous NST events studied in the western Carolinas 
and northeast Georgia.  However, there were several characteristics that 
were somewhat unique to this event when compared to the "distinct 
broken-S" cases that have been previously documented.  In the "distinct" 
cases, the following evolution is typically observed:

A region of high reflectivity (i.e., greater than 50 dBz) within 
a narrow, shallow QLCS exhibits a bulging or bowing appearance.

A narrow channel of lower reflectivity develops across the bulge, 
dividing the QLCS into two segments (northern and southern), with 
the northern extent of the southern segment typically positioned 
downstream of the southern tip of the northern segment (Figs. 
10 and 15).

Tornadogenesis occurs near the southern tip of the northern segment 
as the break develops, or shortly thereafter.

The Bessemer City tornado did not follow this evolution.  With the 
benefit of hindsight, one can argue that a break in the QLCS occurs 
after 0053 UTC.  However, examination of the 0053 UTC reflectivity 
(Fig. 17) reveals that the two high reflectivity segments that were 
later identified as constituting a "broken-S" pattern existed prior to 
the "break" in the line, with the apparent "break" occurring within an 
area of lower reflectivity between the two segments.  This made pattern 
recognition difficult in real-time.  In addition, the southern segment 
did not extend to the north and east of the northern segment.  It is 
interesting to note that while several "distinct broken-S" signatures 
were observed in radar data on this day, apparently none of them were 
associated with a tornado.  

The rear-to-front, descending jet in the Bessemer City storm may have 
played an important role in tornadogenesis.  However, its role was 
probably much different than that of the rear-flank downdraft in 
supercell tornadogenesis.  The numerical modeling studies of Weisman and 
Trapp (2003) suggest that the source of intense, low-level rotation in 
QLCSs is due to downward tilting of environmental, crosswise vorticity 
by sinking air currents, including the rear-inflow jet associated with 
organized mesoscale convective systems.  Since this is the case, the 
amount of environmental, storm relative helicity is inconsequential to 
non-supercell tornado occurrence, because it is a measure of the magnitude 
of streamwise vorticity.  The magnitude of the vertical shear is a more 
accurate measure of the potential for NSTs.  The QLCSs modeled by Weisman 
and Trapp depicted development of strong low-level mesovortices with 
0-2.5 km shear values of 20 m s-1.  The magnitude of the 0-2.5 km shear 
vector calculated from the sounding in Fig. 6 is approximately 26.1 m s-1.  

5.  Concluding Remarks and Operational Considerations 

Detection of non-supercell tornadoes is one of the more difficult 
challenges facing operational forecasters, especially in low-CAPE 
environments, because they do not develop in the classic "top-down" 
manner that has been well-documented in association with supercell 
tornadoes.  It is clear from operational experience that the mechanisms 
responsible for non-supercell tornadogenesis are different from that 
of supercell tornadoes, as NSTs are not associated with mesocyclones.  
Since SRH is an indication of the tendency of thunderstorm updrafts to 
tilt environmental vorticity to the vertical (and form a mesocyclone),
it may have little utility in forecasting the potential for NSTs.  Due 
to the small nature of these tornadoes, their circulations are almost 
always too small to be resolved by the radar beam.  Only in the strongest 
cases is tornadogenesis preceded by an intense circulation in radar 
velocity data.  Even in these cases, radar-indicated rotation will 
likely be detected only within several thousand feet of the ground and 
will precede tornadogenesis only by a matter of a few minutes, if at all.

Acknowledgements

All reference maps were created with Delorme Street Atlas USA 2006.  
The surface analysis and NCEP reanalysis graphics (Figs. 2 and 8) 
were obtained from the Hydrometeorological Prediction Center.  The 
convective outlook graphic, 500 mb analysis, and mesoscale analysis 
graphics (Figs. 3 and 5) were obtained from the Storm Prediction Center.  
The regional surface plot and upper air sounding graphics (Figs. 4 and 6) 
were obtained from the archives at Plymouth State University.  The 
damage survey was conducted by Rodney Hinson.  Larry Lee provided a 
critical review of the manuscript.

References

Grumm, R. H., and M. Glazewski, 2004: Thunderstorm types associated with the
     “broken-S” radar signature. Preprints, 22nd Conf. on Severe Local Storms,
     Hyannis, MA, Amer. Meteor. Soc., CD-Rom, P7.1.

McAvoy, B. P., W. A. Jones, and P. D. Moore, 2000: Investigation of an unusual storm
     structure associated with weak to occasionally strong tornadoes over the Eastern
     United States. Preprints, 20th Conf. on Severe Local Storms, Orlando, FL, Amer. 
     Meteor. Soc., 182-185. 

Thompson, R. L., R. Edwards, J. A. Hart, K. L. Elmore and P. Markowski, 2003: Close
     proximity soundings within supercell environments obtained from the Rapid
     Update Cycle. Wea. Forecasting, 18, 1243–1261.

Weisman, M. L., and R. J. Trapp, 2003: Low-level mesovortices within squall lines and    
     bow echoes. Part I: Overview and dependence on environmental shear. Mon.
     Wea. Rev., 131, 2779-2803. 

Trapp, R. J., and M. L. Weisman, 2003: Low-level mesovortices within squall lines and    
     bow echoes. Part II: Their genesis and implications. Mon. Wea. Rev., 131, 2804- 
     2823.