Mark A. Rose
National Weather Service
Old Hickory, Tennessee
Tennessee does not lie in what is known as the "tornado alley" of the Southern Plains, but its geographical location still allows for a relatively high frequency of tornado occurrences. Since 1830, 469 individual tornadoes that have occurred in Middle Tennessee have been catalogued (Middle Tennessee Tornado Database). In the past decade, during which time the National Weather Service has increased its emphasis on documentation and storm surveys, Middle Tennessee has averaged nearly 16 tornadoes annually.
For purposes of this paper, "Middle Tennessee" is defined as the thirty-nine counties that comprise the county warning area of the National Weather Service Office in Old Hickory, plus three additional Tennessee counties which are covered by the National Weather Service Office in Huntsville, Alabama, but are still considered part of Middle Tennessee (Figure 1).
Troutman and Rose (1997) published a severe weather climatology of Middle Tennessee, which included a section on tornadoes. However, this project only included tornadoes contained within the National Climatic Data Center (NCDC) database, which contains no reports prior to 1950. This climatology expands upon the previous work by Troutman and Rose by including the 109 tornadoes that occurred prior to 1950, as well as the 126 events not covered by the 1997 study (1997-2003).
In this study, the tornado data are sorted into many different categories, including tornadoes by F-scale, hour of day, month of the year, path length vs. F-rating, fatalities per tornado by F-scale, and fatalities per tornado by decade. Separate rankings of the largest tornado outbreaks, longest individual path lengths, deadliest and costliest storms, and tornadoes by county will also be presented and discussed.
2. The F-scale
This study relies heavily on the F-scale in presenting tornado intensity data (Fujita 1971). The F-scale has been reproduced here in order to provide ready-reference (Table 1).
|Scale||Wind Estimate (mph)||Typical Damage|
|Light damage. Some damage to chimneys; branches broken off trees; shallow-rooted trees pushed over; sign boards damaged.|
|Moderate damage. Peels surface off roofs; mobile homes pushed off foundations or overturned; moving autos blown off roads.|
|Considerable damage. Roofs torn off frame houses; mobile homes demolished; boxcars overturned; large trees snapped or uprooted; light-object missiles generated; cars lifted off ground.|
|Severe damage. Roofs and some walls torn off well-constructed houses; trains overturned; most trees in forest uprooted; heavy cars lifted off the ground and thrown.|
|Devastating damage. Well-constructed houses leveled; structures with weak foundations blown away some distance; cars thrown and large missiles generated.|
|Incredible damage. Strong frame houses leveled off foundations and swept away; automobile-sized missiles fly through the air in excess of 100 meters (109 yds); trees debarked; incredible phenomena will occur.|
Please note that F-scale winds are estimates according to observed damage, and have never been scientifically verified. Different wind speeds may cause similar-looking damage from place to place, even from building to building. Without a thorough engineering analysis of tornado damage in any event, the actual wind speeds needed to cause that damage are unknown.
Accurate and consistent assignment of the F-scale is of prime importance both historically and climatologically, however, there are certain deficiencies in this rating system that must be addressed. Although ratings are now conducted by qualified National Weather Service meteorologists, "ratings can vary depending on the experience, background knowledge, and time spent by the person doing the survey..." (Guyer and Shea 2003).
It is also important to remember that the F-Scale rating is based solely on the most extreme damage found along the damage path (that may indicate only a momentary increase in intensity). If the most extreme damage is never seen or reported, the F-Scale rating may be unreliable.
A study by Gordon et. al. (2000) discussed tornadic storms which occurred prior to the implementation of the F-scale. His study discussed that Dr. Schaeffer at the National Severe Storms Forecast Center hired more than 50 college students. One student was hired per state, except Texas, which required several students. Students used such data as photographs, newspaper articles, and information from emergency management. Since information prior to 1950 is typically more "sketchy," F-Scale ratings should be considered more suspect than those determined during the latter 20th century.
3. Tornadoes in Middle Tennessee
The Middle Tennessee Tornado Database was compiled using several sources. Thomas Grazulis' (1993, 1997) exhaustive catalog Significant Tornadoes was used to develop Middle Tennessee's pre-1950 tornado climatology, with most subsequent reports taken from the NCDC Storm Events database.
a. Tornadoes by F-Scale Rating
Sorting the dataset -- which contains 467 rated tornadoes -- according to F-Scale shows a peak of activity under the F1 and F2 classifications, within which a full 69% of tornadoes have been rated. Predictably, the frequency of tornadoes beyond the F2 classification declines rapidly (Figure 2). Only fifty-eight F3 tornadoes are known to have occurred in Middle Tennessee, with twenty-two tornadoes having achieved an F4 rating, and only one F51.
When only tornadoes prior to 1950 are considered, an interesting trend in tornado detection and documentation is revealed (Figure 3). Weak tornadoes (F0 and F1) are almost non-existent, while the distribution of F2+ storms retains similar characteristics to the entire database. That no F0 tornadoes and only one F1 tornado (out of 106 rated storms) were reported prior to 1950 are likely the result of at least three factors.
First, the population across Middle Tennessee was significantly lower, and far more rural. This would increase the likelihood of tornadoes, especially weaker ones, passing undetected. Second, the Weather Bureau prior to 1950 did not issue tornado warnings, and the warning coordination and awareness efforts of the present were not in place then. Therefore, the process of pursuing documentation on tornadic activity was minimal when compared to the processes today. Third, storm damage may not have been considered newsworthy unless it was very significant.
To further investigate this trend, let us now consider tornadoes only within the past decade (Figure 4). The distribution across the F-scale is now far more even, much heavier toward the lower end of the scale, and therefore a more realistic portrayal of the actual distribution of tornado occurrence.
The dramatic increase in detection and documentation of weaker storms is likely a reflection of population increases, greater awareness by the public and media, and the installation of Doppler radar, which is able to detect tornadoes that occur in sparsely populated regions, thereby prompting National Weather Service employees (and the media) to actively pursue storm reports. In fact, 157 tornadoes, or more than one-third of the entire database, have been catalogued within the last decade alone.
b. Tornadoes by Hour of Day
Of the 464 tornadoes reported with a time, most tornadoes in Middle Tennessee, not surprisingly, occur during the late afternoon and evening (Figure 5). The most common hour of the day for tornadoes is between 1700 and 1759 Central Standard Time (CST). More than half (58%) of tornadoes touch down during the seven-hour period between 1400 and 2059 CST. Although tornadoes are possible at any time of the day or night, they are least common during the early daylight hours of 0700-0959 CST.
c. Tornadoes by Month of the Year
When sorted by month of the year, Middle Tennessee's peak tornado season of March, April, and May becomes obvious, as shown in Figure 6. Nearly two-thirds (66%) of Middle Tennessee's tornadoes have occurred during these three months. Again, tornadoes have occurred in every month, but are least common during September, during which only four tornadoes (less than 1%) have occurred. Also note that a secondary peak in activity is centered around November, during which thirty tornadoes (6% of all events) have struck, including a significant outbreak in 2002. Another off-season tornado outbreak occurred in January, 1997 (Table 2).
Again, the data presented in table 2 are obviously skewed toward later years, given the overall increase in tornadoes that are reported. That only one of the aforementioned outbreaks occurred prior to 1974 -- and none prior to 1909 -- does not in any way suggest that tornado outbreaks are recent phenomena, but merely reflects the trend in tornado documentation.
d. Average Tornado Path Length vs. F-Scale Rating
An analysis of tornado path lengths yields typical results. (Path lengths were available for 442 of the 469 tornadoes in the Middle Tennessee Tornado Database, and include the entire track for each tornado -- even portions of tornado tracks that may have occurred outside of Middle Tennessee.) Predictably, average path length increases with F-rating (Figure 7). The average path length for all tornadoes is approximately eight miles. However, this figure increases dramatically beyond F2-rated storms. The average path length for F3 tornadoes, for instance, is 18.7 miles, and increases to 27.3 miles for F4 storms. The only F5 tornado in the database produced a 62½-mile path (Table 3).
|Benton, Stewart (also Gibson, Carroll, and Henry Counties in West Tennessee)|
|Wayne, Lawrence, Giles, Maury|
|Sumner (also Allen, Barren, Monroe, and Metcalfe Counties in Kentucky)|
|Stewart (also Carroll and Henry Counties in West Tennessee)|
|Perry (also McNairy, Chester, Henderson, and Decatur Counties in West Tennessee)|
|Williamson, Davidson, Wilson|
|Benton, Humphreys (also Gibson and Carroll Counties in West Tennessee)|
|Lincoln (also Lauderdale, Limestone, and Madison Counties in Alabama)|
|Lincoln (also Limestone and Madison Counties in Alabama)|
|Giles, Marshall, Rutherford|
e. Tornado Fatalities
Although tornado-related fatalities are now relatively infrequent due to the National Weather Service's warning system and its persistent storm awareness efforts, Middle Tennessee has suffered its share of deadly storms in past years. Typically, as the F-rating increases, so does the potential for fatalities (Figure 8). Throughout the database, there have been 411 deaths attributed to tornadoes. Eighty-six percent of these fatalities have been caused by F3+ tornadoes, which in reality constitute just 17% of all tornado occurrences. And more than half of all fatalities (55%) have been caused by the twenty-two documented F4 tornadoes, which represent 4% of all tornado occurrences (translating to an average of more than 10 fatalities per event). In addition, the four deadliest tornadoes in Middle Tennessee's history were rated F4 (Table 4).
|Sumner (also Allen, Barren, Monroe, and Metcalfe Counties in Kentucky)|
|Giles, Lincoln (also Limestone County, Alabama)|
|Wayne (also McNairy and Hardin Counties in West Tennessee)|
|Lincoln (also Limestone and Madison Counties in Alabama)|
|Davidson, Wilson, Smith|
|Lewis, Maury, Williamson, Davidson, Rutherford|
|Hickman, Maury, Williamson|
|White, Putnam, Overton|
Of the deadliest tornadoes in the database, none have occurred since the two reported during the 1974 Super Outbreak. All of the remaining eight occurred prior to 1934. This is particularly remarkable, given the increases in population which have occurred since then. Davidson County (Nashville), long the population center of Middle Tennessee, received only two of these eight. The city of Columbia (Maury County) saw an F4 tornado on the evening of 20 November 1900 claim 27 lives. The remaining five tornadoes struck primarily rural areas.
To provide direct evidence of the decrease in tornado-related deaths, mean fatalities per tornado have been provided for each decade of the last century (Figure 9). During the first decade of the 1900's, for example, tornadoes produced an average of seven fatalities per event. True, this figure is likely inflated due to the number of non-fatal tornadoes that were never documented. Still, the decade saw 109 Middle Tennesseans fatalities as the result of tornadoes. By comparison, a total of 102 fatalities have occurred since 1950 due to tornadic activity, and 54 of these deaths occurred during the Super Outbreak of April 3, 1974.
f. Costliest Tornadoes
No proper tornado climatology would be complete without some mention of the costliest storms to have struck Middle Tennessee (Table 5). Unfortunately, the NCDC Storm Events database does not contain entries prior to 1950, so details of tornadoes prior to this date are sketchy. Therefore, some of Middle Tennessee's most notorious tornadoes are missing from the following list. (For example, the Nashville tornado of March 14, 1933 caused $2.2 million in damage -- 31.1 million in 2003 dollars.)
|Deaths||Injuries||F-Scale||Location or County|
4. Geographical Considerations
Sorting the tornado database geographically reveals a sizeable range of distribution of tornadoes across Middle Tennessee. Tornado occurrence per county ranges from zero in Clay County to thirty-seven in Davidson County (Nashville). However, because the size of Middle Tennessee's counties is also quite variable, ranging from 737 mi2 (Wayne County) to 114 mi2 (Trousdale County), a more equitable method for comparing tornado occurrence among the forty-two counties involves computing events per unit of area. For purposes of this study, the number of tornadoes per 100 mi2 has been tabulated for each county, in order that a truer comparison can be made (Table 6).
Normalizing tornadoes by land area on a county-by-county scale does not erase the obvious inequalities in population density. Thus, even this method of comparison is not devoid of some deficiencies. However, given the constant change in population growth and shifts, it would be nearly impossible to derive a statistical method to geographically display tornado frequency that is normalized demographically.
In Middle Tennessee, average tornadoes per 100 mi2 ranged from 0.0 in Clay County to 7.4 in Davidson County. The largest such cluster encompasses five counties that include the Nashville Metropolitan Area and areas east (Davidson, Rutherford, Sumner, Trousdale, and Wilson Counties) (Figure 10). True, Davidson County has historically been Middle Tennessee's most populous and urbanized county, greatly lowering the probability of tornadoes passing undetected. However, the maximum in tornado activity across this geographic area spreads eastward into counties that are much more sparsely populated, including rural Trousdale County, which ties its southern neighbor Wilson County with 6.1 tornadoes per 100 mi2 (Figure 1).
Interestingly, of the seven tornadoes in the Trousdale County subset, four occurred entirely within the county. The remaining three were continuations of tornadoes which touched down in counties upstream. Two of the three touched down at nearly identical points (three miles northwest of Lebanon, in Wilson County, which borders Trousdale County to the south). The third tornado touched down 6.1 miles southwest of Nashville, and traveled for 32 miles before lifting in Trousdale County.
The fact that nearly half the tornadoes in Trousdale County originated near the cities of Lebanon and Nashville may have some bearing on the reason this cluster, east of Nashville, exists. Indeed, this cluster may be influenced by the fact that tornado paths that originate near large cities or towns will likely be investigated more intensely and followed to their endpoint (which may be located in more sparsely populated counties). Perhaps, if the same tornadoes had occurred in areas farther away from population centers, they might not have even been reported.
The second cluster is in southern Middle Tennessee, near what is known as the Tennessee Valley of northern Alabama. Specifically, the counties of Giles and Lincoln, which are much more sparsely populated than the northern cluster, are found to have experienced 4.7 tornadoes per 100 mi2, and their northern neighbor, Marshall County, has recorded 5.3 tornadoes per 100 mi2.
When focusing on a subset of F4+ tornadoes, a maximum of occurrence is found to exist in three clusters (Figure 11). The first cluster encompasses a three-county area in southern Middle Tennessee (Giles, Lincoln, and Moore Counties). A second cluster is found over western Middle Tennessee, and includes the rural counties of Benton, Lewis, and Perry. The final cluster is found in northeast Middle Tennessee, and consists of Overton and Pickett Counties, which are also relatively sparsely populated. In fact, Pickett County has the highest frequency of F4+ tornadoes per 100 mi2 in all of Middle Tennessee, with 1.2. Based on this data, there does not appear to be a particular area that is most vulnerable to F4+ tornadoes in Middle Tennessee.
In addition, the subset of F4+ tornadoes may not even be large enough to permit us to draw reliable conclusions. Until a larger data set is acquired, it would probably be improper to make any definitive statements regarding the possibility that one part of the mid state is more prone than another to F4+ tornadoes.
An investigation of the geographic distribution of F3+ tornadoes provides further evidence that strong tornadoes in Middle Tennessee are not generally confined to a particular geographic region. As shown in Figure 12, the number of tornadoes per 100 mi2 ranges from 0.0 to 1.8. There were eight counties in which the frequency exceeded 1.0 tornadoes per 100 mi2. This group includes Giles, Lincoln, and Marshall Counties, but also extends all the way northward to Pickett County on the Kentucky border.
|All Tornadoes||F3+ Tornadoes||F4+ Tornadoes|
The Middle Tennessee Tornado Database contains 469 individual tornadoes occurring during the years 1830-2003. The database has been sorted according to several different variables in order to thoroughly analyze the climatology of tornadoes in Middle Tennessee. Several conclusions have resulted.
1 "The Forgotten F5" hit rural Lawrence County on April 16, 1998, and is the only known F5 tornado to have occurred in Tennessee's history.
The author thanks Henry Steigerwaldt, Science-and-Operations Officer, and Jerry Orchanian, Warning Coordination Meteorologist, both of the National Weather Service Office in Old Hickory, Tennessee, for reviewing this paper. A very special debt of gratitude is owed to Darrell Massie, Lead Forecaster, also of the National Weather Service Office in Old Hickory, Tennessee, for his series of reviews and numerous suggestions which greatly increased the value of this manuscript. Finally, the author extends his appreciation to Bill Conway, Vice President of Weather Decision Technologies in Norman, Oklahoma for his peer review.
Federal Reserve Bank of Minneapolis, cited 2004: Consumer Price Index Calculator. [Available online at http://minneapolisfed.org/research/data/us/calc.]
Fujita, T. T., 1971: Proposed characterization of tornadoes and hurricanes by area and intensity. SMRP Research Paper 91, University of Chicago, Chicago, IL, 42 pp.
Gordon, J., B. Boyd, M. Rose, and J. Wright, 2000: The forgotten F5: the Lawrence County supercell during the Middle Tennessee tornado outbreak of April 16, 1998. Nat'l. Wea. Digest, 24:4, 3-10.
Grazulis, Thomas P., 1993. Significant Tornadoes 1680-1991. Environmental Films, 1326 pp.
Grazulis, Thomas P., 1997. Significant Tornadoes Update 1992-1995. Environmental Films, 118 pp.
Guyer, J.L. and T.J. Shea, 2003: An Assessment of the Variability in Operational Assignment of F-Scale Damage. Preprints, 1st Symp. F-Scale and Severe-Weather Damage Assessment, Long Beach CA .
National Weather Service, Nashville, Tennessee, cited 2004: Middle Tennessee Tornado Database. [Available online at http://www.srh.noaa.gov/ohx.]
Montana State University, cited 2004: Environmental Statistics Group Graphical Locator. [Available online at http://www.esg.montana.edu/gl/xy-data2.html.]
NCDC, cited 2004: Storm Events. [Available online at http://www4.ncdc.noaa.gov/cgi-win/wwcgi.dll?wwEvent~Storms.]
Troutman, T. and M. Rose, 1997: A severe weather climatology for the county warning area of the National Weather Service Office Nashville, Tennessee. NOAA Tech. Memo. NWS SR-190.
Mark Rose is a journeyman forecaster at the National Weather Service Office located in Old Hickory, Tennessee. He previously served as a Navy weather observer (1988-1991) before attending Memphis State University. He graduated in 1994 with a B.S. degree in Geography, including a concentration in Meteorology, and a minor in Mathematics. Mr. Rose joined the National Weather Service at Montgomery, Alabama in May, 1994, and was transferred to Old Hickory in June, 1995. Mr. Rose has authored or co-authored many technical papers, including publications in Weather and Forecasting and National Weather Digest.
Figure 1. Map of Middle Tennessee.
Figure 2. Middle Tennessee tornadoes categorized by F-Scale.
Figure 3. Middle Tennessee tornadoes prior to 1950 categorized by F-Scale.
Figure 4. Middle Tennessee tornadoes, 1994-2003, categorized by F-Scale.
Figure 5. Middle Tennessee tornadoes sorted by the hour during which they touch down.
Figure 6. Middle Tennessee tornadoes sorted by month of the year.
Figure 7. Middle Tennessee tornadoes, average path length vs. F-Scale. (The number above each bar represents the number of tornadoes in each subset.)
Figure 8. Middle Tennessee tornadoes, average number of fatalities per tornado, sorted by F-Scale.
Figure 9. Middle Tennessee tornadoes, average number of fatalities per tornado, sorted by decade.
Figure 10. Counties in which the frequency of tornadoes exceeds 4.6 per 100 mi2.
Figure 11. Counties in which the frequency of F4+ tornadoes exceeds 0.4 per 100 mi2.
Figure 12. Counties in which the frequency of F3+ tornadoes exceeds 1.0 per 100 mi2. The only isolated county in this high frequency group is Pickett County.