Arizona firefighters killed: Would on-site forecasting have averted tragedy?

The local National Weather Service forecast office was supplying forecasts the day the Arizona firefighters died, but on-site forecasters – something used for the most complex firefighting efforts – hadn't been called in.

Was it for want of a weather forecast that 19 firefighters were lost?

As friends, family, and even strangers mourn the loss of all but one member of a "Hotshot Crew" fighting the Yarnell Hill fire, a team of investigators reportedly has arrived to reconstruct the events that led to the deaths of the firefighters Sunday.

The fire, which started last Friday following a lightning strike, had covered only about 200 acres when high winds on Saturday fanned the blaze, expanding it to 2,000 acres within a few hours, according to the Associated Press.

By Monday morning, the fire had enlarged to 8,400 acres, its rough size Tuesday as it burns through grass and chaparral. The fire reportedly has destroyed some 50 homes.

"The winds were coming from the southeast, blowing to the west, away from Yarnell and populated areas. Then the wind started to blow in. The wind kicked up to 40- to 50-mile-per-hour gusts, and it blew east, south, west – every which way," Prescott, Ariz., city councilman Len Scamardo told the AP. "What limited information we have was there was a gust of wind from the north that blew the fire back and trapped them."

The question for meteorologist Cliff Mass, at the University of Washington in Seattle, is whether anyone had pointed out in advance the strong likelihood that this would happen. The downdraft, known as a gust front, came from an approaching storm and served as a blacksmith's bellows, quickly reversing the wind direction that had prevailed through much of the afternoon and suddenly turning the fire back onto the firefighters.

"We have satellite images every 15 minutes and radar every six minutes," says Dr. Mass, a meteorology professor at the university who performs weather forecasts for firefighters in the Pacific Northwest. "This was not a difficult forecast."

Balloon-based measurements of the atmosphere taken at Flagstaff, Ariz., at 1 p.m. local time Sunday showed that in terms of the potential for powerful downdrafts should thunderstorms build, the atmosphere was loaded for bear.

A remote weather station about five miles from where the fire started and set up to monitor conditions in potentially fire-prone areas tells the story from the local perspective, Mass says.

From midday through 5 p.m., winds were blowing generally out of the south at about 30 miles an hour, driving the fire away from Yarnell. At 5 p.m., the wind abruptly shifted, blowing out of the north at about 40 miles an hour with gusts up to 65 miles an hour. At the same time, measurements of sunlight had been falling in an uneven, stair-step pattern since about 2 p.m., indicating that clouds were spreading over the region.

This led Mass to look back over satellite photos and regional radar, which at 1 p.m. showed a line of thundershowers moving toward the area from the northeast. By 3:30 p.m., clouds had moved over the fire site. By 4:45 p.m., a pyrocumulus cloud – a towering cloud formed by rising heat from the fire – had punched through the clouds the storm brought. From above, that was a telltale sign the fire had exploded, Mass explains.

Warning signs also were appearing in the output of the high-resolution, rapid-response forecast model from the National Weather Service (NWS), Mass notes. It put the storm over the fire site about the same time that clouds began to cover the area.

The model is designed to provide fresh 18-hour forecasts once an hour. Another forecast model, at the University of Arizona, yielded similar results.

"That's what bothers me so much," Mass says. "Perhaps this didn't have to happen. The data was there to have saved these people."

Typically, on major fires, incident managers will call on the local NWS forecast office for weather-forecasting support, notes Heath Hockenberry, who manages the NWS's fire-weather program, based in Boise, Idaho.

Incident managers take a look at the complexity of the initial fire and the resources available and determine what people and equipment to send in for the initial attack on the fire, he says. Those decisions include determining if on-site forecasters are needed or whether firefighters can get by relying on the general forecasts that the local forecast office provides.

When forecasters are called in, "we show up with a laptop, a global broadband unit to get satellite data, and usually we just plug into the team's network," he says. "We start forecasting as soon as we get there."

The forecast teams primarily look for anything that has the potential to compromise safety, Mr. Hockenberry says. Thunderstorm outflows, such as those that apparently turned the fire on the firefighters at Yarnell Hill, top the list.

Typically, NWS forecasters are called in for on-site forecasts when a firefighting effort becomes most complex, relying on resources from multiple agencies or political jurisdictions, he says.

That day, the local forecast office was supplying forecasts "that we do every day for every department," Hockenberry says. "Unfortunately, what happened over the weekend was that we didn't have a person on site; we weren't requested."

The fire hadn't reached a level of complexity to prompt incident managers to call in the weather cavalry.

"That's the really sad part," he adds. "There's just not enough [forecasting] bodies to go around to respond to everything. The firefighters – they have to respond to everything."