The Washington PostDemocracy Dies in Darkness

A great forecast but a deadly tornado tragedy in Alabama. What went wrong?

Aerial video from Beauregard, Ala., captures the aftermath of tornados that hit the area on March 3. (Video: Robert Ray, Adriana Usero/The Washington Post)

Despite Sunday’s violent tornado in Lee County, Ala., being well forecast, it marked the deadliest twister in nearly six years. Twenty-three people were killed, more tornado deaths in one day than have occurred nationwide in the past two years.

Some are calling it a meteorological success story, as forecasters identified the risk of severe storms in the Southeast days in advance and provided accurate warnings. Others say it exposes shortcomings in warning systems.

Alabama tornado survivors reckon with ‘Armageddon’

But the infrastructure of the affected area, lacking sturdy structures that could withstand such a force and serve as shelters, has emerged as a primary culprit for the devastation.

We spoke with atmospheric and social scientists alike to get a better picture of what went right and wrong, and the lessons that lay in the rubble.

Reviewing the forecast

By all objective standards, the National Weather Service’s forecast was close to a slam dunk. Meteorologists began to sound the alarm about a severe weather threat as early as last Thursday, zeroing in on the area around Lee County more than 24 hours in advance.

Inside the Southeast tornado swarm that devastated Alabama on Sunday

Bill Bunting, chief of forecast operations at the Weather Service’s Storm Prediction Center (SPC), called the storm system one “with a fair degree of predictability.”

“In the cool season, it’s a bit easier to figure out which areas are at risk farther in advance,” he said. “Then we gradually hone in.”

On Sunday, SPC added a special “mesoscale” forecaster to its team, tasked to analyze the storm environment and pay particular attention to the attributes of individual cells that could produce tornadoes.

More than an hour before the EF-4 tornado packing 170-mph winds dropped, SPC warned that “tornadogenesis will likely occur within the next 30-60 minutes with the possibility of a strong tornado occurring.”

The local National Weather Service office in Birmingham issued a tornado warning 10 minutes before it struck Beauregard and 20 minutes before it arrived in Smith Station.

“All told, I’m pleased with our performance, on how we did with the information we had,” Bunting said. “But we’re always learning from every event, and always striving to improve.”

(For a detailed timeline of the Weather Service’s advisories and warnings for this event, see the bottom of this article.)

Why wasn’t the warning successfully acted upon?

With the red flag flying for days, it’s easy to wonder what went wrong. It’s a case the weather enterprise will undoubtedly be studying for years.

“A perfect forecast is imperfect if not utilized,” Marshall Shepherd, a professor of atmospheric sciences at the University of Georgia, wrote in an email. “The challenge with weather messaging is that human perception and interpretation have taken a central stage alongside model accuracy.”

In other words, it doesn’t matter how good a forecast is if the public isn’t able to understand and act on it.

On Friday, confidence was low that conditions would gel and be ripe for tornadoes. TV meteorologists emphasized damaging winds as the main concern, with tornadoes a secondary threat. They upped the ante by Saturday, but on the weekends television viewership drops significantly.

Although timely warnings were issued, a number of factors stood in the way of the public receiving them. Sirens and TV fail when the electricity is severed as fierce weather approaches, while the former are unreliable anyway. But whether they receive the alert is only half the battle.

“We need to put ourselves in the shoes of the people we’re serving,” Kim Klockow-McClain, a social scientist at the Cooperative Institute for Mesoscale Meteorological Studies, wrote in an email. “We need to understand their circumstances, and the forces that shape their ability to respond. Do they have options for keeping themselves safe?”

In most cases on Sunday, the answer was no.

Walker Ashley, an atmospheric scientist at Northern Illinois University, attributes much of that to building type. “A paramount issue is the large number of mobile, manufactured, and other weak-framed housing throughout that region,” he said in an email. “When even weak tornado affects this type housing, the results can be disastrous – as we witnessed Sunday.”

But change won’t come overnight: In Alabama, 14.7 percent of all households are mobile homes. Lee County’s median household income was just $47,564 in 2017. That’s $14,000 less than the national average. About 17.8 percent of county residents live in poverty, nearly 50 percent more than the national average.

Flimsy housing makes storm targets vulnerable. On top of that, population growth in the Deep South is adding targets and making direct hits more likely. Ashley co-wrote a study in 2016 projecting a significant uptick in observed tornado impacts in the region by the end of the century.

“Average annual tornado impacts . . . are projected to be as much as 6 to 36 times greater in 2100 than 1940,” the study said. That could lead to “the potential for greater tornado disaster potential in the future.”

The study team also emphasized the need for stricter building codes. That could take years. For now, many have no place to seek shelter. A warning is no good if there is no action that can feasibly be taken.

“We must find a way to empower folks living in these housing types, so that they can have the information and resources to take action when a quick fuse warning occurs,” Walker wrote. “We can’t just tell these people to take shelter and hope things are going to go well.”

Although there are no easy answers, better construction practices are a start. “Improving building codes and code enforcement should be thought of as a long-term investment in all of our communities,” Walker wrote. Whether officials take up that torch remains to be seen.

If the storms had roared through on a weekday instead, he said, the fatality count probably would have been much lower; more people would have been at work in well-constructed office or industrial buildings. Instead, people were at home and had their guard down. “This only increases the likelihood of casualties,” he said.

In the meantime, meteorologists are bracing for another dangerous weather system Saturday. Alabama and the southern Mississippi River Valley are again at risk. Sunday’s episode underscores the importance of always having a severe weather plan in place before dark clouds loom. You may have only seconds to act.

Walker warns that the danger won’t wane anytime soon in this part of the world, saying: “We’ve been loading the tornado disaster dice in this region for a long time.”

Storm advisory timeline

3 days, 12 hours before: The Storm Prediction Center outlines portions of Alabama as having a risk of “strong-severe organized convection” on Sunday in its Day 4 outlook.

2 days, 12 hours before: SPC places Lee County, Ala., under a “slight risk” of severe weather on Sunday, where “organized/rotating storms” are anticipated.

1 day, 13 hours before: SPC mentions “a tornado or two may be possible.”

1 day, 2 hours before: SPC upgrades Lee County to an “enhanced risk” for Sunday’s storms. “A potential for strong tornadoes will exist,” it says, highlighting Lee County in the zone where a “significant tornado” may form.

14 hours before: SPC broadens enhanced risk over southern Alabama and into much of Georgia. It forecasts tornadoes, including that “a couple of these tornadoes could be strong.”

7 hours before: SPC mentions a “risk of significant tornadoes.”

4 hours before: SPC issues a mesoscale discussion, emphasizing the chance of strong tornadoes in south and east-central Alabama and western Georgia. The center predicts “rapid environmental changes” and says a tornado watch will soon be issued.

2 hours 30 minutes before: A tornado watch is issued and includes Lee County. SPC calls for a “high” chance of tornadoes and a “moderate” chance of one or more significant tornadoes, and cautions that “a couple intense tornadoes [are] possible.”

1 hour 45 minutes before: SPC issues a special discussion regarding “a maturing supercell along the GA/AL border near Columbus” and cautions that “a couple of tornadoes are possible during the next few hours with this activity.” It is unusual for SPC to issue a discussion centered on a single storm cell.

1 hour 15 minutes before: SPC again zeroes in on a cell near Montgomery just to the west of Lee County, and states that “tornadogenesis will likely occur within the next 30-60 minutes with the possibility of a strong tornado occurring.”

23 minutes before: The National Weather Service in Birmingham issues a tornado warning for “Central Lee County … [and] Northwestern Russell County.”

14 minutes before: The NWS updates the tornado warning, tracking “a confirmed large and extremely dangerous tornado” and describing it as a “PARTICULARLY DANGEROUS SITUATION!”

12 minutes before: An extremely rare tornado emergency is declared.

The tornado strikes between 2:20 and 2:25 p.m.