The EF scale: how an EF5 rating is assigned

The Enhanced Fujita scale rates a tornado from the damage it left behind, not from any wind anyone measured: a surveyor matches what is on the ground to one of 28 damage indicators, reads an estimated wind speed off that indicator, and the tornado’s rating is the category that estimate falls into. An EF5 is assigned when the damage at some point along the path supports an estimate above 200 mph.

The tornadoes on radar guide covers what a radar shows while a tornado is happening. This page covers the number that arrives afterwards, produced by an entirely different process.

What the original Fujita scale got wrong

Ted Fujita published his scale in 1971 to separate weak tornadoes from strong ones, and both the meteorological and engineering communities adopted it almost immediately. Its wind speeds were expressed as the fastest quarter-mile wind: F0 at 40–72 mph rising to F5 at 261–318 mph, as the Storm Prediction Center still lists them.

The Texas Tech Wind Science and Engineering Center’s recommendation report to the National Weather Service names three limitations: “a lack of damage indicators, no account of construction quality and variability and no definitive correlation between damage and wind speed.” Its engineering objection is blunt: “a 260 mph wind speed is not required to completely destroy a well constructed house and blow away the debris,” citing Minor et al. (1977) and Phan and Simiu (2003) on F4 and F5 housing damage overestimating the winds needed to cause it.

The scale that replaced it

The Texas Tech project, led by James McDonald and Kishor Mehta, grew out of a forum of 23 scale users held in Grapevine, Texas, on 7–8 March 2001. Its recommendation reached the NWS in 2004 and was revised twice, most recently on 10 October 2006. The scale became operational in the United States on 1 February 2007.

The report derived those numbers: it converted Fujita’s quarter-mile speeds to three-second gusts using the Durst curve, applied a regression between the two scales, y = 0.6246x + 36.393 with R² = 0.91, and rounded the results “to avoid implying more accuracy than justified.” That is why an F3 becomes an EF3 even though the two categories carry different wind numbers.

EF5 was deliberately left open at the top: “an absolute upper bound on tornado wind speed has not been defined to date,” and leaving EF5 unbounded “will prevent the news media from always assuming the worst case scenario.” SPC is equally direct about the whole table — “The Enhanced F-scale still is a set of wind estimates (not measurements) based on damage” — and notes the three-second gust “is not the same wind as in standard surface observations.”

28 indicators, and the degrees of damage

The 28 indicators run from small barns through houses, manufactured homes, motels, strip malls, schools and warehouses to transmission line towers, free-standing poles, and hardwood and softwood trees. Each carries its own ordered list of degrees of damage, from the threshold of visible damage to total destruction, and each degree carries three numbers: an expected wind speed with a lower and an upper bound. Three of the ten degrees for a one- or two-family residence show the ladder.

The bounds exist because construction varies. The expected value assumes “no glaring weak links, traditional construction quality, appropriate building materials, compliance with local building code and continuous maintenance.” Two nails where three belong push the estimate toward the lower bound; hurricane clips push it up. That range is where most disputed ratings actually get decided.

The speeds themselves came from expert elicitation, not measurement. Six experts — two meteorologists, two engineers, an architect and one person trained in both fields — estimated the expected, lower and upper bound speed for every degree of damage under a protocol borrowed from seismic hazard assessment, refined over three rounds.

A single building never decides an event. NWS guidance says surveyors identify the indicator “from more than one of the 28 used in rating the damage,” and the rating represents the highest wind anywhere in the tornado’s life. The scale also has a ceiling: the upper bound for total destruction of a house is 220 mph, and “the actual wind speed in the tornado could have been higher since there is not another DOD to indicate a higher wind speed.”

The rating depends on what was standing there

A damage scale can only rate damage, so a violent tornado crossing open pasture rates low. El Reno in 2013 is the case in point: mobile radars sampled winds near 300 mph, the survey found only EF3 damage, and EF3 is what it was rated.

The Enderlin, North Dakota tornado of 20 June 2025 shows how far surveyors go when the standard indicators run out. NWS Grand Forks upgraded it to EF5 on 6 October 2025, 108 days later (our calculation), on forensic wind estimates for tipping loaded grain hopper cars and lofting tanker cars — one empty tanker thrown about 475.7 feet — plus debarked and root-ball-displaced trees along the Maple River. At one farmstead the foundation was swept clean, but “issues surrounding the lack of proper anchoring will limit the overall rating in this area.”

How rare an EF5 is

SPC’s list of the most intense damage category holds 60 tornadoes rated F5 or EF5 in the United States since 1950. Ten of those fall after the EF scale took effect, from Greensburg, Kansas on 4 May 2007 to Enderlin on 20 June 2025.

Against the whole record the proportion is small. SPC’s database for 2007–2025 holds 24,162 records, of which 22,616 carry a rating; the rest are logged as unknown, a category that has grown since 2016. Ten of the rated tornadoes are EF5: 0.044%, about one in 2,260. EF4 and EF5 together account for 108, or 0.48%. Just over half of all rated tornadoes are EF0. (All four figures are our calculations from the SPC file.)

The gap between EF5s is the part worth understanding. From the Moore, Oklahoma tornado of 20 May 2013 to Enderlin, 4,414 days — just over twelve years — passed with no US tornado rated EF5 (our calculation from SPC’s list). Violent tornadoes did not stop: the database records EF4s in every year from 2013 to 2025 except 2018. Whether that gap reflects meteorology or survey practice is not something this page can settle. SPC cautions that official NWS ratings “may differ on occasion from those of other tornado databases,” and 2025 figures remain preliminary.

What radar contributes, and what it does not

Radar cannot assign a rating, and not because of any particular radar’s limits. A Doppler radar measures one component of motion along its beam, averaged over a sample volume, above the ground; an EF rating is an inference about a three-second gust at a damage point on the surface. After El Reno, the NWS stated that “EF ratings are determined by observed damage rather than measured wind because we have no consistent way to measure wind speed for every tornado that occurs.”

Radar still contributes, in a supporting role. It corroborates: NWS Grand Forks noted that Enderlin’s peak estimate “correlates to the maximum strength on WSR-88D Storm-Relative Velocity data from KMVX” — radar agreeing with a survey rather than replacing it (see radar velocity). It points surveyors at the path: the tornado debris signature marks where debris was airborne, and its height and extent track intensity loosely — NWS warning guidance treats a signature reaching 8,000–10,000 feet as pointing to EF2 or stronger, and Van Den Broeke and Jauernic (2014) found signature area increasing with EF rating. And it fills gaps at night or in rain-wrapped storms nobody could see into.

The limits are specific. Radar cannot see the surface at range, so the winds that decide a rating sit below the beam. It cannot tell a debarked tree from an intact one, or a house that failed at 120 mph from one that failed at 190. And it answers in minutes where a rating takes days to months.

WeatherStory Radar reads full-resolution Level II, including velocity and correlation coefficient, straight from NOAA’s public archive.