The short answer is that a tornado is almost never measured at all. It is rated afterwards, from what it destroyed. Almost every wind speed you have seen attached to a tornado is an estimate inferred from wreckage rather than a reading taken from an instrument, and understanding why explains most of the arguments that surround tornado records.
Why can't you just measure the wind?
Because anything capable of taking the reading is usually destroyed by the thing it is reading. A standard weather station anemometer is not built to survive 200 mph, is not likely to be standing in the right place, and is measuring at a fixed height on a mast rather than at the level where the damage happens. Tornadoes are small, short-lived and move, so the probability that one passes directly over working instrumentation is very low. Direct measurements exist, but they are rare enough to be individually famous.
What was the Fujita scale?
Ted Fujita introduced his scale in 1971 at the University of Chicago, working with Allen Pearson, and its insight was to work backwards. Rather than measure the wind, survey the damage and infer the wind speed that must have caused it. The scale ran F0 to F5, and crucially each rating was a band of wind speeds rather than a single figure. An F4 meant 207 to 260 mph. An F5 meant 261 to 318 mph. This is why older records in any archive, including this one, often appear as a range: that is the rating, faithfully reproduced, not a hedge.
What is the Enhanced Fujita scale?
The Enhanced Fujita scale became operational in the United States on 1 February 2007. Canada adopted it in April 2013, with 31 damage indicators rather than the 28 used in the United States. The reason for the change was that Fujita's original wind estimates turned out to be too high: buildings fail at lower speeds than he had assumed, so the old scale was systematically overstating the winds behind severe damage.
The EF scale keeps the same six-point structure but reworks the wind figures and formalizes the survey. It uses 28 damage indicators, categories of thing that can be damaged, running from one and two-family residences through schools, towers and transmission poles to hardwood and softwood trees. Each indicator carries a list of degrees of damage, each with an expected wind speed range attached. A surveyor identifies what was hit, judges how badly, and reads off the estimate. EF5 begins above 200 mph.
What is measured, exactly?
The figure the scale assigns is an estimated three-second gust at the point of damage. That last part matters and is routinely dropped. It is not a sustained wind speed, it is not the fastest instantaneous velocity anywhere in the vortex, and it is not measured at the standard ten-metre height a weather station reports from. The National Weather Service is explicit that these speeds vary with height and exposure and are not directly comparable to weather station readings. When a mobile radar reports a much higher number for the same tornado, a large part of the gap is that the two are describing different things in different places.
Where does radar come in?
Mobile Doppler radar, most famously the Doppler on Wheels, can measure rotational velocity directly, and has produced the highest wind speeds ever recorded on Earth. But under operational National Weather Service rules those readings do not set the rating. A radar velocity is taken well above the surface, samples a volume rather than a point, and exists for only the tiny fraction of tornadoes a research team happened to be pointing at. Blending them into a record built from damage surveys would make the record internally inconsistent.
The 2013 El Reno tornado is the case where that policy became a public argument, and we have covered it separately in why the El Reno tornado was not rated EF5.
What about path length and width?
These are surveyed rather than estimated, which makes them the most reliable numbers attached to any tornado. A damage survey establishes where the circulation first reached the ground and where it lifted, and the width is the maximum extent of the damage swath. They are recorded in the Storm Prediction Center database alongside fatalities, injuries and loss, and they are the figures this archive ranks on wherever they exist.