Why forecasting tornadoes is so difficult

It isn't unusual to have tornadoes in winter, especially in the Gulf Coast states, but the number of tornadoes that have struck the southern and central United States over the last week have been unusually numerous and especially deadly.

A heavily damaged area in the aftermath of a tornado in Rowlett, Texas, Dec. 27, 2015.
Laura Buckman | AFP | Getty Images
A heavily damaged area in the aftermath of a tornado in Rowlett, Texas, Dec. 27, 2015.

"In fact it is the deadliest December since 1953, when tornadoes killed 49 people," said Bill Bunting, operations chief for the Storm Prediction Center at the National Weather Service.

A more typical December might bring a few tornadoes over the course of one or two days, Bunting said. Speaking with CNBC on Monday, he said tornadoes have struck those regions every day for the previous seven days — an exceptionally long streak.

Of course, other storms, such as hurricanes, frequently destroy property or threaten lives. But one thing that sets tornadoes apart is how difficult they are to predict.

Tornadoes are most common in the Southeast and the Central Plains, but they have been reported in all 50 states. They can, and often do develop suddenly and quickly, and the exact conditions that create a tornado are still a bit of a mystery. Sometimes conditions that could create a tornado, do not.

"We don't fully understand every aspect of tornado development, and while we have (observation) systems that are very good, we don't know everything about the atmosphere at a particular location at a particular time," Bunting said.

The strongest tornadoes tend to form in the middle of rotating thunderstorms called "supercells." In "tornado alley" in the the Midwest, tornadoes seem to be influenced by the mixing of warm air coming up from the Gulf of Mexico and cooler air drifting down from the northwestern region of the country.

"While there are some days that we know are going to produce storms ... there are other days where conditions that are favorable for a tornado to form come together so briefly — for a few hours and over a very small area," he said. "You really have to be incredibly diligent and never assume that it can't happen today because conditions don't look favorable. All it takes is one relatively small change in how the atmosphere is structured or how it is behaving, and that can make an incredible difference in the outcome."

That means that the National Weather Service often has little time to issue a proper warning — sometimes only a few minutes — before a storm hits.

The National Severe Storms Laboratory is working on two programs that it hopes will change that. The Warn-on-Forecast program, and the FACETs program (for Forecasting a Continuum of Environmental Threats), are both long-term experiments in using computer modeling to better predict possible storm outcomes.

Right now, meteorologists have to rely primarily on Doppler radar or on information from people spotting storms in order to issue tornado warnings.

The WoF and FACETs programs would allow researchers to enter a variety of information — satellite, radar and other data — and run scenarios determining the likelihood of a tornado hitting a given area.

That could increase warning lead times from little more than 10 minutes to 30, 60, or even 90 minutes, Bunting said. But several years may pass before the projects are ready for actual forecasting.

In the meantime, people need to pay close attention to forecasts and take warnings seriously, he said.

"The atmosphere doesn't have a calendar," Bunting said. "If the conditions are coming together for severe storms and tornadoes, it doesn't matter whether it is May or December."