A tornado that formed this month in Mangum, OklahomaReuters

It has been an exceptional week for tornadoes in the United States. In the past 13 days, more than 366 twisters have struck the central and eastern United States. They ranged from a deadly tornado in Dayton, Ohio, to a mile-wide beast that leveled houses across 30 miles of Kansas but remarkably left no fatalities.

This outbreak has been abiding: While no one day has proved catastrophic, it’s odd that so many medium-caliber tornado days should follow one another so closely. Even on Wednesday, a relatively quiet reprieve from the storms, the National Weather Service warned of tornadoes from Texas to New Jersey.

This stubborn outbreak originated from a stubborn weather pattern. Strong thunderstorms give rise to strong tornadoes—and for the past two weeks, all the features have been in place for midwestern tempests. A hot area of high pressure over Georgia and South Carolina has ferried moisture from the Gulf of Mexico clockwise into the center of the country. Once there, it meets a cold trough of air near the Rockies, where it can rapidly condense. The jet stream passing nearby contributes enough atmospheric instability to whisk all that moisture into a storm. And so a strong thunderstorm—and possibly a tornado—is born.

The United States has a hybrid, almost cyborg method for predicting tornadoes. For decades, meteorologists have been able to detect likely tornadoes remotely by looking for a hook shape on weather radar. But in the past three decades, there have been two major steps forward. First, forecasters can now detect wind speed and direction via Doppler radar, allowing them to identify centers of circulation in storms. Second, they can use reflective radar to look at the presence of objects in the atmosphere that are neither water nor cloud: debris. This technique, made available just in the past few years, lets them confirm the existence of a tornado remotely, by looking for debris lofted high in the sky.

But the service still relies on networks of in-person spotters, first formed in the 1940s as protection for military assets, to confirm the existence of a single tornado. (An earlier Army experiment with tornado spotters and forecasts ended in 1886; a Signal Corps officer had barred the use of the word tornado in the warnings, fearing that it might incite public panic.) On Tuesday, as a mile-wide EF-4 tornado cut across the Kansas City metropolitan area, in-person spotters could affirm its existence in addition to radar.

According to a preliminary report, nobody died in the tornado, though 18 people were injured. If that number holds, it’s “absolutely phenomenal,” Patrick Marsh, a forecaster at the Storm Prediction Center, told me.

This week is the biggest outbreak of tornadoes since the spring of 2011, when hundreds of storms wracked the Plains, killing more than 550 Americans. The death toll was so high, and the damage so awful, that the National Weather Service changed how it communicates severe weather in response. The service now favors direct, plainspoken, and nonscientific language when warning of an incoming tornado, hurricane, or other disaster. On Tuesday, for instance, National Weather Service forecasters explicitly told media outlets in Kansas City to use the strongest language possible when discussing the danger of the massive tornado.

Forecasters’ thinking on the benefits of direct communication has evolved dramatically over time. A century ago, the Weather Bureau, then run by the Department of Agriculture, prohibited its forecasters from mentioning tornadoes to the public at all; in the most dangerous circumstances, they could warn instead of “destructive local storms.” During the decades this ban held, from the early 1900s to the 1950s, research into tornadoes and their formation withered.

The success of the military’s spotter network and its forecasts began to dissolve the U.S. government’s official resistance to talking about tornadoes. In 1948, two military weathermen successfully forecast a tornado in the vicinity of Tinker Air Force Base; the military also developed a habit of leaking tornado forecasts to the public. At the same time, radar was being turned from military to meteorological uses, increasing both the accuracy and utility of tornado warnings. In July 1950, the U.S. Weather Bureau changed its policy. With routine forecasting and directed research, tornado warnings have become reliable and specific enough to create a life-saving buffer for people in their paths.

While this week’s outbreak is enormous, it still pales in comparison with the scale of the 2011 outbreak. Five of the worst 25 tornado days in U.S. history happened in 2011, Marsh said. By contrast, none of the worst 25 tornado days in history has happened this year.

Indeed, there’s no evidence that tornadoes are getting worse in response to climate change, Victor Gensini, a professor of extreme weather at Northern Illinois University, told me. Roughly the same number of tornadoes happen today as happened in the 1950s, when reliable tornado record-keeping began, drawing data from both radar and spotter networks. What is likely to change as the climate warms is the frequency of bad storm years, he said. Tornado activity will become “spikier,” with violent years interrupting relatively peaceful ones, he said.

That’s already borne out in the data, Genisi said. Just last year, scientists were discussing the tornado drought since 2011. Now that period is clearly over.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.