A recent decline in the accuracy of tornado warnings is not as alarming as it may appear, weather officials say.
Forecasters are doing well with storms that produce strong tornadoes – those that measure EF-2 or higher on the Enhanced Fujita Scale, said Russell Schneider, director of the National Oceanic and Atmospheric Administration’s Storm Prediction Center in Norman, Okla.
A review of warnings issued by National Weather Service meteorologists shows that average lead time given to the public on tornado warnings across the U.S. has fallen from 15 minutes in 2011 to nine minutes in 2016. But “almost all that drop is associated with weaker tornadoes,” Schneider said.
In 2011, National Weather Service forecasters were able to issues warnings ahead of a tornado 75 percent of the time. By 2015, that figure was 58 percent.
“Weaker tornadoes, they are difficult to forecast and warn for,” Schneider said.
Because weak tornadoes make up the bulk of the total, he said, that can skew the warning accuracy statistics.
“Is that the only cause” for the drop in warning accuracy? Schneider asked. “There is research looking into it.”
Accurate, timely, effective
Congress is demanding improvements.
The Weather Research and Forecasting Innovation Act of 2017, which was signed into law last month by President Donald Trump, mandates that the NOAA reduce the loss of life and economic losses from tornadoes through the development of accurate, effective and timely tornado forecasts, predictions and warnings. That includes developing the capability of reliably predicting tornadoes at least an hour ahead of time.
The challenge with comparing tornado warning accuracy statistics with past years, weather officials say, is that the basic parameters changed significantly in 2007.
Instead of issuing warnings for entire counties, the weather service identified smaller zones that encompass areas directly in the path of thunderstorms producing – or capable of producing – tornadoes.
The change allowed forecasters to exclude areas within a county not threatened by severe weather, thus reducing the false-alarm rate.
The national rates of false alarms – when a warning was issued but no tornado materialized – were about 76 percent in the 10 years before Greensburg, Hayes said. In the decade since Greensburg, that figure has dropped to about 73 percent.
From 2008 to 2010, the average lead time of warnings for strong tornadoes was 17.5 minutes, Schneider said. In 2011, a record-setting year for tornadoes that included a deadly outbreak in late April and the devastating Joplin, Mo., tornado in May, the average lead time was 19.5 minutes.
But from 2012 to 2017, that figure fell to 15 minutes.
For all tornadoes, the average lead time dropped from 13 minutes in 2008 to nine minutes in 2016.
How much time is too much?
A decrease in warning lead times is not necessarily a bad thing, Schneider said.
“Longer warnings may not be more effective,” he said. “They may be less effective. People may assume they have a lot of additional time to do other things” before seeking shelter.
“Is a 60-minute warning four times better than a 15-minute warning?”
Even before Congress passed the Forecast Innovation Act in April, the NOAA set up a unit called the hazardous weather test bed in Norman. It brings together more than 50 scientists from around the world, forecasters and warning coordination meteorologists to work on new techniques to better warn the public, Schneider said.
The deadly outbreak in April 2011 in the southeastern U.S. prompted weather officials to partner with experts in social science to more closely examine how people receive and respond to weather warnings. That included examining how people perceive false alarms, he said.
Integrated Warning Teams, made up of weather officials, emergency managers and members of the media, have been established to review and improve language used in warnings to more clearly describe the threat and impact of threatening storms.
Along with issuing warnings in small targeted zones, weather officials now have the ability to send weather warnings to smartphone apps inside those zones. No longer do they have to hope residents are at home watching television or within range of an outdoor siren.
“We have a much more nimble society,” Schneider said, so tools needed to be developed to reach them.
Officials in Greenwood County in southeast Kansas directly credited those alerts sent to phones with saving lives when a strong tornado struck Eureka after sunset last July. Residents were warned that a tornado was approaching, and they took cover.
There were no fatalities and only a few injuries despite the tornado striking at night.
“There’s many stories like this,” Schneider said. “The information is getting out to people much more quickly.”
No decline in Kansas branches
No decline in Kansas branches
The dip in tornado warning lead times hasn’t occurred in Kansas, said Chance Hayes, warning coordination meteorologist for the Wichita branch of the weather service.
The average lead time for the four weather service branches in Kansas before the Greensburg tornado was 13.7 minutes, Hayes said. In the 10 years since, it has been 14.26 minutes.
“When you look at the four core offices for Kansas, there was improvement in every statistic that’s kept,” Hayes said.
That includes the false alarm rate, which is important because it means residents are less likely to ignore warnings when they’re issued.
The false alarm rate for Kansas branches in the 10 years before Greensburg was 64.5 percent, Hayes said. Since 2007, it’s been 60.1 percent. Those figures were at least 10 percent lower than the national average.
“When your false alarm rate goes down, that’s good,” Hayes said.
‘Very, very worrisome’
AccuWeather senior vice president Mike Smith has called the decline in tornado forecasting accuracy “very, very worrisome.”
The warning accuracy rate hasn’t been as low as 58 percent nationally since 1994, Smith said. He attributes that in part to the retirement of many experienced meteorologists in recent years.
“In a lot of cases, they don’t have the training and the experience” to more accurately predict tornadoes, Smith said of current meteorologists.
But Hayes said he doesn’t share Smith’s concern about the recent decline in warning accuracy statistics.
“The offices here in the heartland perform extremely well,” Hayes said. “We’re very educated, very well-trained staffs across the whole country.”