By Roscoe Nunn, Meteorologist.
US Weather Bureau.

(Delivered before the Tennessee Academy of Science, Nov. 25, 1921, being the Annual Address of the retiring President)

(Transcribed by Mark A. Rose)

The first large contribution to the statistical knowledge of tornadoes in America was made by Lieutenant J. P. Finley, of the old Signal Corps of the Army, which was the predecessor of the Weather Bureau. Lieutenant Finley made a thorough study, in the eighteen-eighties, of the character of six hundred tornadoes, records of which he collected. These 600 tornadoes occurred during the long period, 1794 to 1881. Finley also made an elaborate report on the tornadoes of the year 1884, which was a most unusual year for these storms, more than 60 tornadoes occurring on one day, February 9, of that year.

Before and after the investigations of Finley, several contributions to the literature of the subject were made by Signal Corps or Weather Bureau Officials, notably Professor William Ferrel, who gave the most satisfactory theory of the conditions that cause tornadoes and the most plausible explanation of the tremendous energies developed in these most violent of all storms. He gives a detailed discussion of the subject in his "Popular Treatise on the Winds," published in 1889. It seems that the principal features of Ferrel's theories are still acceptable to meteorologists, and in giving a brief explanation of the phenomena of tornadoes, I shall endeavor to follow Ferrel's treatises.

Statistical data of tornadoes have, of course, been kept up by the Weather Bureau since the days of Finley and Ferrel, and articles have been published from time to time, giving valuable information, but explanations of their cause, which according to Prof. W. J. Humphreys, must be purely mechanical, have been mainly restatements of Ferrel's ideas. And so, in this paper, I cannot hope to contribute anything new or original except some information in regard to the occurrence of tornadoes in Tennessee, which I shall give before I close.

Tornadoes are almost entirely an American product, though they do occur occasionally in a modified form in other parts of the world. They are not only confined very nearly to the United States, but occur principally in certain portions of the country, and not in all portions. They are almost unknown in the Rocky Mountain districts, or west of the 105th meridian, and they very rarely occur in the Allegheney Mountain sections. They seldom occur along the Gulf and Atlantic coasts and seldom north of latitude 45. The area of greatest frequency comprises, roughly, Missouri, Kansas, and Iowa, but they occur frequently enough also in northern Alabama, Georgia, South Carolina, western Tennessee, eastern Arkansas, Illinois, Indiana, Ohio, and southern Michigan.

I have made a few charts showing the distribution of tornadoes in the United States during recent years, and will show them here on the screen.

(Slides for years 1916, 1917, 1918, and 1919.)

Now, as to why tornadoes are peculiar to certain regions of the United States. In a word, it is due to geographical conditions. The tornado districts of the United States happen to be the most suitable in the world for the assembling of meteorological conditions out of which tornadoes develop. The position and trend of the Rocky Mountains, with the Great Plains to the eastward and southward, induce a strong southward flow of air from the northern regions under certain barometric conditions, while the location of the Gulf of Mexico from which winds blow northward with the passing of low pressure areas eastward over the central valleys, makes possible a great northward movement of warm, moist air. Often, therefore, strong countercurrents are to be expected in the Mississippi Valley regions, and when the running of a cold current from a northerly direction over a warm current from a southerly direction occurs, there results a combination of meteorological conditions that may produce tornadoes.

So long as the cold wind passes under the warm, there will be no great disturbance, for the equilibrium of the air will be stable; but, if the warm advances under the cold, the equilibrium becomes unstable, of course, and if the contrasts are strong enough, tornadoes may result.

It seems that the conditions just described seldom obtain in any part of the world other than the tornado regions of the United States, for in no other part are the geographical conditions so favorable for them.

Tornadoes occur most frequently in the spring and early summer. Prof. A. J. Henry compiled data for an eight-year period and found that during the eight years there were 6 days in January on which tornadoes occurred; in February, 10; March, 16; April, 31; May 42; June, 51; July, 25; August, 11; September, 4; October, 2; November, 3; December, 5. But it must not be assumed that the same relative frequency obtains each year. Occasionally there are more tornadoes in April than in any other month, yet on an average of a number of years it is probable that there will be more in May and June than in any corresponding period. These data refer to the United States as a whole.

It is found that in the winter months, tornadoes occur in the Gulf States, but very seldom in other regions; that with the gradual warming up of the valleys and plains of the interior they occur farther and farther to the northward until the month of June, which is the month of greatest frequency in Nebraska, South Dakota, Iowa, and Minnesota.

It is further shown that the tornadoes of April and May frequently occur almost simultaneously over very considerable areas, while those of the summer and winter months are, as a rule, restricted to the smaller boundaries of a single State. Also, that the very destructive tornadoes occur principally in the late spring and early summer months, March to June, inclusive.

In Tennessee, during the last 33 years, April has a greater number of tornado dates than any other month, viz., 6; then comes March, with 5, May with 4, January with 2, and February, July, August, September, October, November, and December*, with 1 each; while June has no tornado accredited to it during the last 33 years. On some occasions in Tennessee as many as four different tornadoes occurred, almost simultaneously, travelling in nearly parallel paths. So that, with a total of 26 different dates for the 33 years, we have a total of 47 tornadoes.

Tornadoes cause considerable loss of property in the United States in the course of the average year. Prof. Henry found that in an eight-year period the total loss was $23,767,200; but in that period was included the St. Louis tornado, the most destructive to property on record, showing a loss of over $13,000,000. The Louisville tornado, with a property loss of nearly $3,000,000, was also included. The St. Louis and Louisville tornadoes are of course very rare instances. Most tornadoes cause only small losses, owing to their short and narrow paths. It is estimated that what may be classed as very destructive ones occur on an average about three times a year in the United States.

Lightning causes many times more damage to property and kills many more people than tornadoes.

Yet, the loss of life in tornadoes is sometimes appalling. There are several instances in which 100 to 300 people were killed in a single day. There is a record of a tornado in Adams County, Mississippi, May 7, 1840, in which 317 people were killed, and another in the same county about two years later, June 16, 1842, in which 500 were killed. The St. Louis tornado, May 27, 1896, killed 306. On the afternoon of March 28, 1920, at least 13 tornadoes occurred, eleven of them in Illinois, Wisconsin, Indiana, and Michigan, and two in Alabama and Georgia. These storms killed 163 people. But the most memorable date in tornado history probably is that of February 9, 1884, when a series of tornadoes occurred from Mississippi, Tennessee, Kentucky, and Illinois, eastward to Virginia, North Carolina, South Carolina, and Georgia. On that day there were more than sixty separate tornadoes after 10 a. m. Over ten thousand buildings were destroyed, 800 people killed and 2500 wounded.

(Slides 942, 959, 939, and 930). Showing destructive effects.

Now, as we have seen where and when tornadoes occur, and have had some information as to their effects, it may be interesting to give a description of a typical tornado and tell something of the theory of its origin.

First, let us see a photograph of a tornado cloud. (Slide 933).

The funnel-shaped cloud is the special distinguishing feature of a tornado as viewed in the air. The funnel-shaped cloud is the center of action. The cloud may or may not reach the earth. If it touches the earth it leaves a trail of destruction; if not, little or no effect is produced. Little or no destruction is wrought on either side of a rather clearly defined path. The width of the path may be only a hundred or two feet, or it may be a half mile; it averages about 1000 feet. The length of the path varies from a few hundred feet to many miles; occasionally it may be 300 miles long. The tornado nearly always travels northeastward, but sometimes changes its course considerably; sometimes it moves towards the southeast or east. Its rate of travel differs greatly, but averages about 25 miles an hour, while its rotary winds have a velocity estimated at not less than 100 miles an hour in mild form and as high as 500 miles an hour when most violent. No other storm wind approaches this in violence. The roar of the tornado is said to be terrifying.

The tornado is a storm within a storm; that is, it is always associated with a cyclonic system, or low pressure area. It always develops in connection with a thunderstorm, but it is different from the thunderstorm proper. The thunderstorm is a "straight wind" storm, while the tornado, as its Spanish name implies -- that is, "turning" -- has rotary winds, and is marked by a funnel-shaped cloud.

The tornado usually occurs in the southeastern quadrant of an area of low pressure, some distance from the center of the low, probably where there is the greatest conflict between opposing masses of air that have sharply contrasted temperatures.

Contrasts in temperature and the flowing of counter currents cause turbulence, especially at levels a few hundred feet above the ground, and when it happens that a great body of warm, moist northward-moving air is overrun by a great body of cold, southward-moving air, there results an unbalanced situation that must soon right itself in a great overturning, probably with violent upward currents. Sometimes these vertical currents starting from the ground break through the upper cold layer in many places, giving rise to a number of tornadoes almost simultaneously.

The upward currents take a spiral, counterclockwise course, the same as the winds of the cyclone under whose influence they occur, showing their definite relation to the cyclonic system, or low pressure area, which may be from 500 to 1000 miles in diameter.

Now, if at some place of updraft, or outlet, countercurrents are drawn into the same rising column, they may be so deflected as to augment the incipient whirl into a most violent state. The law of the conservation of areas is effective here, and the rising currents are brought in spirally with increasing angular and linear velocity as the axis of the spin is approached, and thus, near the center, almost circular winds with velocities reaching 100 to 500 miles an hour are produced, which results in a partial vacuum at the center of the whirl. And so, the funnel-shaped cloud forms. It is caused by condensation of the moisture of rising air, as it cools rapidly by expansion in the partial vacuum.

Then, if the funnel-shaped cloud establishes good contact with the ground, destruction begins. It is as if some tremendous air-pump were being carried along a few hundred feet above the ground, sucking up the air over which it passes. This accounts for the apparent explosion of houses and closed vessels over which the tornado cloud passes. The pressure being suddenly relieved from the outside, the houses and vessels are bursted by the sudden expansion of the air within. It is well known that if atmospheric pressure were reduced suddenly, say, one-half, it would cause a pressure of about 7 pounds per square inch in the inside surfaces of all objects containing air that could not quickly escape.

An interesting thing about the paths of some tornadoes is the fact that they are broken here and there with stretches that were left unharmed. We sometimes see it said that "the tornado travelled with a bounding motion," or, "the tornado lifted, then came to the earth again." These expressions may carry the idea that the tornado as a whole rises and falls. But the idea is suggested that the upper portion of the tornado may be travelling on a horizontal line -- not bounding -- but that if the whirl produces a partial vacuum down to the earth so that large obstacles are encountered, the resistance of such obstacles will after a time have the effect of breaking up the lower end of the vacuum, thus shortening the funnel-shaped cloud (which would then appear to have risen); that, after a period of smooth running of the whirl aloft, the vacuum again becomes more powerful and is again developed down to the earth. The process may be repeated until finally the partial vacuum is so broken up by contact with trees, buildings, and hills, that it is dissipated entirely; or else the atmospheric conditions may be so changed that the tornado cannot be sustained. Then the tornado is spent.

There is not time to speak of the marvellous pranks of tornadoes, which we have all read about; such as the driving of nails head first into planks, shooting cornstalks into wooden walls, stripping harness from horses, demolishing a house but leaving some article unmoved while one that stood within a few inches of it disappeared entirely, carrying people high in the air and setting them down again without serious injury, etc.

So we will now show a few slides illustrating the relation of tornadoes to low pressure areas and the effect on the barograph record of a nearby tornado.

(Slides 1400, 1450, 1708, 1709, and 1780.)

(Maps and diagrams explained)

We will then take up a brief discussion of Tennessee tornadoes, with charts showing their paths, etc.

Tennessee Tornadoes.

We have taken the tornado records for Tennessee during the last 33 years for this study. This period will be sufficient for the purposes of the paper, which are to show the general distribution of tornadoes in Tennessee, the direction of movement, approximate length of paths, etc.

The charts which we will show on the screen present the facts in such manner as to make extended remarks unnecessary.

But it will be worth while to speak of a few Tennessee tornadoes that were specially noteworthy on account of the unusual destruction wrought, length of paths, etc. As in other parts of the country, so in Tennessee, most tornadoes are not very destructive on account of the very small area they cover.

On November 20, 1900, the State was visited by four tornadoes during the afternoon and evening hours, two or three of which proved quite disastrous to human lives and destructive to considerable property.

(Attach extract from "Climatological Data", Tennessee Section, November, 1900.)

During the afternoon and night of November 20, 1900, very severe storms occurred at many places in Middle and West Tennessee, and in several localities they assumed the extreme violence of tornadoes. From all information gathered, there were at least four different tornadoes in the State on that day, between about 3 p.m. and 10 p.m., Central time. They moved in a northeasterly direction. The first reported struck LaGrange, in Fayette County, about 3 p.m.; the second passed over Williamson County and the northern portion of Rutherford County between 6:30 and 6:45 p.m., and the third occurred in Maury County, between 9 and 10 p.m., striking Columbia at 9:30 p.m. In each case the storm was accompanied by terrible clouds, terrific thunder and lightning and heavy downpours of rain.

At LaGrange, three people were killed and about a dozen houses wrecked, including several churches and the Southern Railway depot.

In Williamson and Rutherford counties 9 people were killed and many others wounded - 2 were killed in the neighborhood of West Harpeth, 5 at Nolensville, and 2 at Lavergne. The destruction to property amounted to about $60,000 in the vicinity of Lavergne; in Williamson County it amounted to thousands of dollars, but a careful estimate has not been obtained. The path of this tornado varied in width from about 50 to 300 yards.

At Columbia, and in that vicinity, 27 people were killed and 75 wounded, and the property loss amounted to about $40,000. The tornado cut a swath about 300 yards wide and 3 miles long. On approaching Columbia it swerved a little to the westward and thus missed the main portion of the city. This was one of the most destructive storms in the history of the State.

It is possible that the list of dead given above should be larger by two or three. It is certain that there was considerable damage in localities not mentioned above by the heavy rains and high winds. The streams in Middle Tennessee were flooded in a few hours.

But the most disastrous tornadoes, so far as we have record, in Tennessee, occurred on April 29, 1909, when 60 perople were killed.

(Attach extract from "Climatological Data", Tenn. Sec., April, 1909.)

The first violent winds seem to have occurred about 2 p.m. in Hamilton County. This disturbance apparently moved northeastward and struck the Knoxville section about 4 p.m.

The most clearly defined and the longest tornado track was made across the State from the extreme southwest corner northeastward into Scott County. This tornado came from Mississippi, struck White Haven, Shelby County, Tenn., at 7:30 p.m.; struck a section a few miles east of Somerville, Fayette County, at about 8:30 p.m.; vicinity of Henderson, Chester County, about 9 p.m., vicinity of Centerville, Hickman County, about 10:30 p.m., Hillsboro and Franklin, Williamson County, about 11 p.m., Algood and Cookeville, Putnam County, about 1 a.m. (30th). The width of this track varied from a few hundred yards to about one mile.

A second well defined tornado track extended from the vicinity of Cuba, Shelby County, northeastward through Haywood, Gibson, Carroll, Humphreys, Dickson and Montgomery counties. This track was parallel with the one first described and the time of occurrence of the tornado was about the same. Its width varied from about 200 yards to about two miles.

A third track extended through Giles and Lincoln counties and northeastward, the tornado being apparantly dissipated in the mountains. It struck the vicinities of Pulaski, Bryson and Fayetteville about midnight. Its track was about one-half mile wide as it passed through the southern part of Giles County.

A fourth track passed through Franklin County, the tornado striking Decherd about 12:30 a.m. April 30 and apparently passing with diminished violence through Grundy and Meigs counties. Its track was about two miles wide at Decherd.

Each of these terrific disturbances was accompanied by the usual distinctive marks of the tornado - pendent funnel-shaped cloud, narrow path, destructive violence and scattering of debris in all directions.

Deaths were reported as follows: Giles County, 22; Lincoln County, 7; Williamson County, 8; Montgomery County, 4; Chester County, 4; Hickman County, 10; Franklin County, 4; Hardeman County, 1. The list is not complete for the State. A hundred or more people were reported injured, a number fatally. The property loss was estimated as follows: Giles County, $100,000; Lincoln County, $100,000; Montgomery County, about $25,000; Williamson...

It should be rememered, however, that the probability of any particular place being visited by a tornado is very slight, even in the regions where tornadoes occur most frequently. For example, Prof. Cleveland Abbe made careful calculations some years ago, which showed that the probability of any area of 1 square mile being struck by a tornado is less than one-sixteenth of one per cent per century.

The question may be asked whether Nashville was ever visited by a tornado. It seems that none of much consequence has struck Nashville. It is possible that a storm of February 12-13, 1880, was a light tornado. This passed over the city about midnight and damaged the Custom House, "moving several blocks of granite and tearing down a gable end."

On March 25, 1884, a light tornado occurred about 6 miles north of Nashville, striking a part of the Gallatin Pike.

On May 27, 1917, one passed just north of Brentwood in an east-northeast direction, through Bakerton and Una, in Davidson County, about 8 miles southeast of Nashville.

(Insert extract from "Climl. Data", May, 1917)

Between Perry and Davidson counties we have no report of any great violence. It is probable that after passing Linden it was deflected upward by the western Highland Rim, which extends in a north-south direction across Hickman and Lewis counties. This is assumed from the fact that it was not so violent in these counties, but became more destructive on reaching the eastern part of Davidson County, where it seemed to come to the earth again, passing just north of Brentwood, in an east-northeasterly direction, through Bakerton and Una, thence through Dodoburg and Lebanon, Wilson County. It passed from Brentwood to Lebanon, a distance of about 30 miles, in 45 minutes.

In the Bakerton-Una district two lives were lost and about 30 persons injured. At Dodoburg several persons were injured and many houses and much timber destroyed, including a large poplar grove valued at $30,000. The storm passed through a portion of the residence section of Lebanon, partially wrecking many homes and uprooting trees, but its force was much spent by this time and the damage here was relatively slight. No persons were reported injured at Lebanon. The total property loss in Davidson County is estimated at $75,000, and in Wilson County $85,000.

If it be correct that this was the storm that originated in McNairy County, it traversed a distance of about 140 miles between 6:00 p.m. and 8:45 p.m., or at an average rate of approximately 50 miles an hour.

As the tornado passed near Nashville (about eight miles distant), the station records show its effect decidedly. A thunderstorm had been in progress since 6:45 p.m., but was very mild until about 8 p.m., when the wind and rain suddenly became furious. At 7:55 the wind velocity was 20 miles per hour, at 8:00 p.m., 33 miles, and from 8:03 p.m. to 8:05 p.m., 72 to 74 miles an hour, decreasing to 12 miles an hour at at 8:10 p.m. The maximum rainfall amount in 5 minutes, .56 inch, equalled the highest previous five-minute record at this station, which occurred on November 21, 1900, at the time of the passing of a tornado about 30 miles south of Nashville.

The following circumstances are reported in connection with the several tornadoes:

The Cooperative Observer at Dyersburg reports: "Immense hail on 27th." It fell between 4 and 5 p. m.

Near Dyersburg a gypsy camp was completely demolished, the occupants saving themselves by climbing into concrete culverts. It is reported that only one horse survived and it was carried a distance of 500 yards.

Bruce Bryant, whose brick residence near Trezevant was wrecked, reports that the sun was shining when the storm passed him (5 p.m.), that the main cloud appeared to be not over 150 yards high, and that there was no rain with the storm. The members of his family took refuge in a ditch and were uninjured.

Several coats were found in the southeastern part of Henry County that had been blown a distance of more than 20 miles. They were identified as belonging to persons who were killed near Trezevant.

An interesting account is given by Miss Maggie Moran, whose home was wrecked at Una, Davidson County: " * * * the door was blown open; I tried to shut it but couldn't, and about that time the roof and north wall were blown northward and I was hurled to the floor, only a second, and the wind turned and blew the rest of the kitchen away and I was blown fifteen feet from the kitchen in a southeasterly direction. One foot was caught between two logs, which were supposed to come from a tree which was blown down near me; the shoe on the other foot has never been found."


Losses and Casualties by Tornadoes in Tennessee, May 27, 1917.

Cates Landing
Dyersburg and vicinity
Sharon (near)
Ore Spring
Trezevant (near)
Indian Mound
Center Point (near)
Scott's Hill

NOTE. -- Charts** were then thrown on the screen, showing tornadoes in Tennessee, years 1900, 1909, 1917, and 1921; also a chart showing all tornadoes for a period of 33 years, 1889-1921, incl.

Explanations and running comments were made.

This closed the address.

* One occurred Dec. 23, 1921 (after this paper was presented), entering Tipton County from Arkansas.
** These slides sent to C.O., with those borrowed from C.O. for this lecture. Also sent to C.O. in the same package four slides made at Nashville, showing tornadoes for 1916, 1917, 1918, and 1919, for whole country, reproduced from C.W.B. Annual Reports.


List of all tornadoes of record for a period of 33 years, 1889 to 1921, inclusive.

Year Date County Place Time
1889 ...None...
1890 March 27 Sumner near Rogana 8:30 p.m.
May 10 Shelby near Memphis 7:30 p.m.
1891 ...None...
1892 April 4 Weakley Sharon 8:30 p.m.
May 10 Shelby, Fayette near Collierville 12:45 p.m.
1893 March 23 Madison near Jackson 5:00 p.m.
Weakley Dresden 5:30 p.m.
April 12 Hardeman Middleton 2:00 p.m.
Haywood Brownsville 4:00 p.m.
April 14 Chester near Henderson 4:00 p.m.
1894 ...None...
1895 January 21 Tipton Covington 6:00 a.m.
July 6 Gibson Idlewild 8:30 p.m.
1896 ...None...
1897 ...None...
1898 ...None...
1899 August 26 Weakley near Gleason ?
1900 November 20 Fayette, Hardeman La Grange 4:30 p.m.
Williamson, Davidson, Rutherford Nolensville-LaVergne 6:00 to 6:30 p.m.
Hardin, Wayne, Lewis, Maury Savannah, Columbia 7:00 to 9:30 p.m.
Fayette Moscow 7:15 p.m.
1901 ...None...
1902 ...None...
1903 ...None...
1904 ...None...
1905 ...None...
1906 ...None...
1907 ...None...
1908 ...None...
1909 April 29-30 Shelby northeastward to Scott White Haven, near Somerville, near Henderson, near Centerville, near Franklin, near Cookeville 7:30 p.m. to ~1:00 a.m.
Shelby, Haywood, Gibson, Carroll, Humphreys, Dickson, Montgomery 7:30 p.m. to ~midnight
Giles, Lincoln, Bedford, Coffee Harms, Ardmore Midnight
Franklin, Grundy, Sequatchie, Bledsoe ~12:30 to ~2:00 a.m.
October 14 Madison, probably Chester & McNairy, Hardin Denmark, Stantonville, Pittsburg Landing 4:45 to 5:20 p.m.
Wayne Clifton ( 7 to 10 mi. SE) ~6:00 p.m.
Giles Lynnville (3 mi. S) 6:00 p.m.
1910 ...None...
1911 ...None...
1912 ...None...
1913 March 13 Hardeman, McNairy, Henderson, Benton Bolivar, Finger, Huron. Life, Lexington, near Camden ~12:30 to ~3:00 p.m.
Giles, Maury Bodenham, Culleoka, Rallyhill 3:30 p.m. to ?
Giles, Marshall, Bedford Brick Church, Cornersville, Eagleville, Smyrna 3:30 p.m. to ?
1914 ...None...
1915 May 6 Crockett, Gibson, Carroll, Henry Alamo, Atwood, Trezevant ~8:15 p.m. to 10:00 p.m.
1916 January 12 See Annual Report C.W.B.
1917 March 23 See Annual Report C.W.B. & Tennessee Reports
May 27 (4)
October 29
1918 ...None...
1919 September 21 See Annual Report C.W.B. & Tennessee Reports
1920 April 20 (4) See Annual Report C.W.B. & Tennessee Reports
1921 February 16 See Reports
March 24
April 16 (4)
May 10 is the U.S. government's official web portal to all federal, state and local government web resources and services.