(This article, first published Thursday afternoon, was updated Friday morning.)

New research could help reshape how we think tornadoes form.

The study, presented Thursday at the American Geophysical Union annual meeting in Washington, expands on previously published work that investigated the massive El Reno tornado in 2013. It had winds up to 300 mph, and was the widest tornado ever observed. It killed several storm chasers. And it is one of the most-studied tornadoes of the past five years.

When Jana Houser and her colleagues looked back at the data, they noticed something they hadn’t observed before — instead of starting in the clouds and dropping to the ground, the tornadoes they examined appeared to have started at the ground and worked their way up. The 2013 El Reno tornado was among these. In all, they found five cases in which their radars detected that signature. It’s a small sample size, but it could prove invaluable to the severe-weather research community.

The idea that tornadoes form, or at least appear to form, from the ground up is not a new theory. What’s significant, according to scientists not involved in the study, is that this could be the start of a database that could be used to prove one of several hypotheses on tornado formation.

“It is important to have these case studies to build evidence for particular theories,” said Karen Kosiba of the Center for Severe Weather Research.

The research is based on analysis of data from a network of mobile radars deployed by the University of Oklahoma and other institutions. National Weather Service radars, which can see storms up to about 120 miles away, cannot detect signatures at ground level.

“You actually have to tilt the antenna up in order to make the beam go over contaminating objects [such as trees and hills] toward the ground,” Houser said.

Mobile radars, however, are driven toward tornado-producing storms to collect vital data on their wind and structure at the lowest levels of the atmosphere. Houser said they “were deployed on a slight hill that enabled us to collect data.”

The results back up one of the favored theories about “tornadogenesis,” or the formation of a tornado. In this theory, the conduit is the a rush of cooler air that blows from the storm toward the area where a tornado develops, which meteorologists refer to as the rear-flank downdraft. This feature is common in supercell thunderstorms, where, far above the ground, the entire storm is rotating. Not all rear-flank downdraft surges lead to tornadoes, but they have been observed as one of the signs a tornado might be imminent.

The most significant challenge in tornado research is not coming up with theories, but having the observations to support them. The severe-weather research field is so limited in data because of the short nature of the storms under study. Without months-long missions employing a dozen meteorologists and the most advanced mobile radar technology, observing a tornado develop and recording that data is next to impossible.

“We don’t know beforehand when these storms are going to produce tornadoes,” Houser said at a news conference Thursday. “Being in the right place at the right time is challenging.”

"We have relatively few high-quality case studies of tornado formation,” she added, and now these four tornadoes are among them.

Victor Gensini, an extreme-weather scientist not involved in the research, said he is supportive of the research and the way the group is gathering the data.

“We need challenges to our conceptual model of how things work,” Gensini said, though he cautioned against drawing major conclusions from this particular study. Even so, he concludes, “it’s novel in how they’re approaching the problem with the radar observations.”

The research may ultimately help forecasters issue more reliable tornado warnings. Presently, only ten to 20 percent of rotating thunderstorms produce tornadoes, which leads to many false alarms.

“If we can really figure out how tornadoes are formed, this will give us better insight into why some storm produce them while others (that are in the same environment!) don’t,” Houser said in an email.