The duck boat that sank in Missouri's Table Rock Lake in July 2018 is raised. (Nathan Papes/Springfield News-Leader/AP)

It has been 18 months since a fully loaded duck boat plunged beneath turbulent whitecaps atop Table Rock Lake near Branson, Mo.

The ill-fated tour took place despite warnings of dangerous weather. Less than 40 minutes later, the vessel had been swallowed by angry waves, extinguishing 17 lives as 70 mph winds raked the lake above.

A National Weather Service assessment report released in December is highlighting the dangers of “self-interpretation” of weather data in the wake of the tragedy. This is in reference to situations in which non-specialists or untrained individuals make high-stakes weather-related decisions.

The assessment, which was compiled by seven Weather Service employees and a number of social scientists, also examined best practices for communicating weather information to vulnerable outdoor populations.

The original forecast


High-resolution National Weather Service radar shows the outflow boundary of strong winds arriving well ahead of any rain, lightning or thunder. (NWS)

At first glance, it may appear that what happened July 18, 2018, was a perfect forecast falling on deaf ears. The Storm Prediction Center had issued a severe thunderstorm watch at 11:20 a.m. advertising a “high” chance of severe wind; at 6:32 p.m., a severe thunderstorm warning was issued for Table Rock Lake. The warning, which mentioned damaging winds, came 28 minutes before the onset of destructive gusts and 36 minutes before the sinking. The boat hadn’t entered the water until 6:55 p.m., 23 minutes after the warning to seek shelter came out.


Storms formed in Kansas around 10:30 a.m. on July 19, 2018, about eight-plus hours before affecting the Table Rock Lake area in Missouri. (NOAA/GR2 Analyst)

According to the report, the boat’s captain “reviewed radar information” between 6:25 and 6:27 p.m., six minutes before beginning the tour. Before 6 p.m., the general manager of Ride the Ducks Branson “reviewed radar information … and made a self-assessment on timing of the derecho approaching.”

Challenges in the assessment

Dick Wagenmaker is the meteorologist in charge at the National Weather Service office in Detroit. He also was the team lead tasked with spearheading the service assessment. Wagenmaker said the group’s initial goal was to deconstruct how decisions were made by the duck boat company based on weather information the team accessed.

“Our original intent was to examine how the duck boat operators used publicly available information,” he explained.

But ongoing litigation got in the way; specifically, nobody from the Weather Service was able to interview anyone from Ride the Ducks Branson, which has been embroiled in lawsuits since shortly after the disaster.

“From that, we changed our primary focus … [to] how vulnerable populations use and respond to [Weather Service] products and services,” the assessment said.

Balancing forecaster confidence and public thirst for information


Flowers for the victims are placed at the ticket counter at Ride the Ducks Branson in July 2018. (Charlie Riedel/AP)

The findings of the report centered on two key themes: disseminating weather warnings and the content of what’s in the warnings themselves.

The team found that emergency management officials, third-party partners and other planners alike prefer weather warnings in the pathcast format. That means a warning about dangerous weather with a list of locations and a predicted time of arrival.

Emergency managers and “other users were almost unanimous in [preferring] pathcast” Wagenmaker said.

But that’s at odds to what the meteorologists constructing the warnings often default. Forecasters frequently use a locations impacted format, which lists various landmarks in a warning bulletin that are in a storm’s predicted path — without the down-to-the-minute estimates that a pathcast offers. It just mentions the communities that are likely to be affected during the lifetime of the generally less than one-hour warning.

Obviously, knowing precisely when a storm may strike is advantageous. But forecasters often choose to leave that out because of uncertainty, especially in rapidly evolving environments in which storm speed or behavior is changing. Besides, knowing a storm might strike at 10:04 vs. 10:07 doesn’t change the actions you should take.

Meteorologists are stuck between a rock and a hard place. Many of them don’t want to issue pathcasts. If they don’t, they worry the public will try to do it themselves. That can be a recipe for disaster.

Why forecasters leave out pathcasts

In an ideal world, pathcasts are a good idea. But when they falter, the implications can be deadly.

A 2009 National Weather Service report found that the predicted track of one particular Missouri tornado, based on a computer-generated pathcast, differed by as much as eight miles from the twister’s actual location. That confused emergency managers, who sounded sirens in the wrong part of Newton County, Mo. Twenty-one people died, including a storm spotter, who positioned himself in the tornado’s actual path while thinking he was safe, based on the warning’s pathcast.

That warning came from the Weather Service in Springfield, Mo. The service’s neighboring office in Tulsa has a policy to never issue warnings with pathcasts. Meteorologists there still don’t like pathcasts.

“Including the pathcast assumes a constant linear storm motion,” said Ed Calianese, warning coordination meteorologist at the Weather Service in Tulsa. He said that’s often not realistic in real-life storm environments.

“A lot of time, it’s not that easy,” he said. “It’s a moving target. And each time you [update the warning], all those numbers are going to change anyway.”

Calianese emphasized the importance of acting as soon as the warning is issued, rather than prioritizing exactly what time it will arrive.


A National Weather Service document detailing the sources of error in pathcasts. (NWS)

In a National Weather Service training document used to discuss best practices for weather warnings, 21 pages are dedicated to “the error sources inherent” to pathcasts.

“Errors of 5 miles or so are quite frequent” in specific pathcasts, with timing often five to 10 minutes off, as well.

The document seemed to echo long-standing forecaster sentiment.

“By including exact times and locations we run the risk of implying forecast precision we do not have.”

Emergency management craves precise storm arrival times

Yet emergency managers crave exact arrival times in storm warnings, even though predicting them accurately is sometimes an impossible game. “Our recommendation is to develop additional training to encourage forecasters to use pathcast … and how to keep it fresh and up to date,” Wagenmaker said.

“Forecasters tend to be more cautious, and some forecasters expressed concern about errors in the timing of” pathcasts, he said.

A difficult choice

That leaves a difficult choice. The service assessment recommends that forecasters include specific information about a forecast that can change minute to minute.

“We do know from our interviews that it’s pretty important … to people to get a time of arrival. There’s a tendency that, if it’s not available to [the public], to sort it out themselves.”

In other words, the Weather Service knows that issuing pathcasts is a risky business. But it could be even riskier if the public tries to do it on its own.

“We have no idea if the dispatchers at Ride the Ducks Branson tried to do their own interpretation on time of arrival,” Wagenmaker said.

What could have been missed

Meteorologists don’t just look at the glossy green, yellow and red radar you may see on cellphone apps. There are at least a dozen Doppler radar-derived products used to make judgment calls during severe weather situations.

In the case of Branson, the damaging winds preceded rain by as much as 10 miles or more. Meteorologists spotted that gust-front feature by locating a fine line on radar. But on most commonplace cellphone apps, that subtle feature could easily have been erased by smoothing algorithms.

“There are limitations to many smartphone apps that smooth or filter low reflectivity values from their radar display,” the report stated. Most people, when glancing at a radar display, conjecture that the worst impacts will be felt beneath the brightest colors. In many situations, that’s simply not true — “especially when the leading edge of thunderstorm winds are not coincident with or near reflectivity cores.

“Untrained users may not properly recognize or understand that high winds can be associated with gust fronts … and incorrect estimates of speed of movement can result in incorrect arrival time estimates.”

The bottom line

The report stated the company personnel had, in multiple instances, used radar data they themselves accessed. “The captain made a verbal reference to looking at weather radar as passengers were boarding,” according to the National Transportation Safety Board, at 6:29 p.m. Two minutes earlier, the captain had personally reviewed radar data with his supervisor.

Presumably aware of the impending severe storm, they altered plans to complete the water portion of the tour first.

Between 6:30 and 7 p.m., the nearest cloud-to-cloud and cloud-to-ground lightning strikes were 10.9 and 13.1 miles away from the incident location, respectively; thunder probably was not heard, even as winds approaching hurricane force tore across the lake. This absence of rain/thunder also may have contributed to the decision despite professional warnings urging shelter.

The “use of smartphone apps and self-interpretation of radar imagery … is likely a permanent reality,” wrote the Weather Service. Self-interpretation can “reinforce behaviors that delay sheltering.”

A key take-home implication of the Weather Service’s assessment is the following: It’s imperative to heed all warnings and not delay action for a second opinion or confirmation.

The assessment recommends wireless emergency alerts be pushed out for especially severe thunderstorms to call attention to particularly dangerous situations. At the time of the duck boat disaster, wireless emergency alerts would sound only for tornadoes and flash floods.