University of Arizona meteorologists are using a sophisticated forecasting model, partially developed using UA research, to better understand the weather conditions that played into the tragic events that claimed the lives of 19 firefighters battling the Yarnell Hill Fire southwest of Prescott, Ariz. on June 30.
The UA is home only a handful of atmospheric science departments in the country that run a forecasting model in real time, just like the National Weather Service. The UA version of the Weather Research and Forecast Model, or WRF, offers about five to 10 times more detail than what is commonly available to the National Weather Service, and it is the only one of its kind in the Southwest.
"The following morning, we performed what weather forecasters call a 'post-mortem' – which is what we typically do in the aftermath of severe weather events," said Michael Leuthold, a meteorologist and computer expert in the department of atmospheric sciences
. "We look at the weather data to get a better idea of the conditions leading up to that event and how well the model performed in forecasting it."
Such analyses not only help weather experts better understand the conditions leading up to extreme weather events, but the data are being used to improve the accuracy of short- and long-term weather forecasts.
"There are two weather stations located close to Yarnell Hill, so from a meteorologist's point of view, we have a pretty good idea of what happened that day," Leuthold said.
The data showed 19 mph winds from the southwest, with gusts up to 22 mph – nothing too unusual. But at some point between 4 and 5 p.m., around the time the firefighters went into the area that would become a death trap, winds rapidly shifted to the opposite direction and doubled in speed.
"Thunderstorms had built up that day north of the area, and they were moving south and toward the fire," Leuthold said. "Typically when they start moving down into the low deserts, they die off, but while they do, they send out strong outflows, and those are columns of rain-cooled, denser air that flows out of a decaying thunderstorm. Those flows hit the ground and spread horizontally. In general, those downdraft are relatively benign, but they can also come in violent microbursts, and may have been what happened in this case."
After they had pieced together the meteorological conditions surrounding the event, Leuthold wanted to see whether the model was able to forecast the dramatic weather events that were observed and contributed to the danger of the situation.
"We took the model run from Sunday morning and checked if it would have forecast the wind speed, sudden shift in direction, the location and the timing over a 12-hour forecast period," he said.
To generate a forecast, the model relies on a flurry of data fed into it from weather stations, oceanic sensors and weather balloons, among other sources.
"Our model generates a three-dimensional grid of the atmosphere with parameters like temperature, relative humidity, wind speed, air pressure and so forth at a specific point in time," Leuthold explained. "The model then takes these initial conditions and applies mathematic equations that represent processes in the atmosphere. Essentially it computes forward what the atmosphere will look like as time progresses."
In the analysis of the Yarnell Hill Fire, the meteorologists performed two model runs using slightly different parameters. Leuthold pointed out that these runs were actual forecasts using data that would have been available to the model before the event, with no actual, weather data added after the fact.
"The first one wasn't quite in the right spot, but the second one was able to forecast this quite well."
In a large area of relatively slow wind speeds, the model had put a tiny red dot right on top of Yarnell Hill, symbolizing localized but very high wind intensity with wind speeds in the 40 mph range.
According to the researchers, the sudden shifts in wind direction and speed, in combination with very low fuel moisture and extremely high temperatures set the stage for the deadly events.
"The problem with weather forecasting is the uncertainties expressed as percentages," Leuthold explained. "For example, say there is a 30 percent chance of thunderstorms over the Catalina Mountains. Firefighting doesn't work like that. Firefighters have to make a decision on the spot. They either go in or they don't."
Improving the resolution and accuracy of weather forecasts could dramatically improve the safety of firefighting crews in the future. By constantly incorporating weather data into their models, meteorologists like Leuthold are working to make forecasts better.
"You'd be hard pressed to model a small-scale feature like Yarnell using the National Weather Service model," he said. "They couldn't run a model like this for every area across the nation because there is not enough computing power."
Another advantage of the UA forecasting model is its capability of taking the terrain into account in addition to modeling atmospheric processes, Leuthold said.
"Most of the thunderstorms in the summer are created by the mountain ranges," he explained. "We've tailored the physics in the model to better present weather processes that are specific to our area: from the extremes in temperatures between winter and summer to the changes in humidity from almost zero percent to nearly tropical during the monsoon."
A clever arrangement enables the UA department of atmospheric sciences to maintain such a high-powered forecasting system: The meteorologists teamed up with the research faculty to invest in high-power computers, which are taken over by the forecasters in the morning, while the researchers use them in the afternoon since they rarely need real-time data.
The model was developed and is being improved with input from the UA researchers as part of a collaboration among many partners, including the National Weather Service, the National Center for Atmospheric Research, the U.S. Air Force and other universities across the world. All data and forecasts are publically available on the department of atmospheric sciences
Several partners have contracted with the UA to take advantage of the forecasts. In addition to producing the forecasts for solar and wind energy generation for Tucson Electric Power, for example, other agencies like the Pinal County Department of Environmental Quality and the Maricopa Flood Control District rely on the forecasts. Idaho Power, Idaho's main power company, contracted with the UA to run the Arizona model to generate forecasts for renewable energy, especially wind and hydropower.
"Idaho Power is now using our model for all their forecasts," Leuthold said. "The company sponsors some of our graduate students to improve the model. In their words, it’s the best model out there."