Australia’s Fire Season Ends, and Researchers Look to the Next OneWith its otherworldly conditions, Australia has become a testing ground for fire-prediction technology.
By Helen Sullivan
SYDNEY, Australia — On March 2, for the first time in 240 days, not a single bush fire burned in the state of New South Wales. The state’s Rural Fire Service declared the worst fire season in history, during which 25 people in NSW were killed, officially over. In those eight months, 6 percent, or 13.6 million acres, of the state that a third of Australians call home had been incinerated.
The world’s attention, riveted on the fires earlier this year, has understandably shifted to the ongoing coronavirus crisis. But the devastating fire season has left lessons in its wake. As Australia looks toward a future of more frequent and dangerous fires, scientists and officials are working together to develop fire-prediction technologies that will enable firefighters to work faster and more safely when the next season — expected to be perhaps equally grueling — begins in just a few months.
What Australia continues to learn could be used elsewhere — everywhere from other countries, including the United States, to outer space, in software that must withstand the searing, blustery and otherwise inhospitable conditions of other planets.
When a wildfire breaks out, one of the most difficult decisions faced by the operations team is who and what to send where, and which resources to keep in hand in case they are suddenly needed elsewhere.
“Whether you hold resources back in reserve in case more fires break out, or whether you hit that fire very hard, can mean the difference between a fire that’s put out in 15 minutes and one that goes for weeks,” said Greg Mullins, a former commissioner of Fire and Rescue New South Wales. To make that decision correctly, firefighters first must know which areas are high-risk.
Central to many of the more recent technologies is the ability to predict the influence of Australia’s eucalyptus trees on a given fire. Eucalyptus are particularly fire-intensive; their dry, shedding bark catches easily, and the embers can be blown ahead of a blaze, lighting others. This phenomenon is known as “spotting,” and it is one of the most challenging problems in predicting a fire’s behavior.
An Australian computer program called Phoenix RapidFire models this kind of spotting, simulating the spread of fires across a given area. It has been relied upon to predict fire behavior in both Victoria, where it was introduced after the Black Saturday bush fires that killed 173 people in 2009, and New South Wales. A similar program, FarSite, is used in the United States.
When a wildfire starts, analysts at the NSW Rural Fire Service headquarters in Sydney, who may be 200 miles away or more, enter variables into Phoenix, such as the fire’s location, the time it started and the terrain. Closer to the fire, regional teams feed information back to headquarters, where the fire management team, with the help of manual analysts, decides where to send resources like firefighters, trucks and water-bombing helicopters.
The technology does not yet outperform people when predicting the spread and behavior of a particular fire. Simon Heemstra, the manager of planning and predictive services at the NSW Rural Fire Service, who has a Ph.D. in fire behavior, described Phoenix as mainly a “triage tool.”
“Nine times out of 10,” he said, manual analysts produce more accurate results than the model. Using their experience, analysts are able to incorporate the uncertainty inherent in fire behavior, something “the computer just isn’t able to grasp.” But where the computer model excels, Dr. Heemstra said, is in analyzing several fires at once and determining which one poses the greatest risk — and therefore which one manual analysts should focus on.
Australia’s national science agency, the Commonwealth Scientific and Industrial Research Organization, has developed computer software called Spark, which aims to improve upon Phoenix.
Phoenix was built to predict fire behavior in forest and grass, Dr. Heemstra said, so for several other fuel types, like shrub land, “it’s a bit like trying to fit a square peg in a round hole.” Spark, because it uses unique equations for each fuel type, is more intuitive and reliable. It could be “the next evolutionary step” in firefighting models, Dr. Heemstra said, and the NSW Rural Fire Service hopes to use it as early as the next fire season.
Whereas fire behavior models like Phoenix and Spark help predict the spread of a fire, drone technology may be able to predict where fires are likely to start. For the moment, drones are used mainly to monitor grassland fires. Forest fires burn particularly hot, and are volatile, making them unsafe for drones to fly over or for anyone nearby to operate the devices.
The wildfire conditions in Australia are sufficiently severe that they verge on otherworldly. NASA’s Jet Propulsion Laboratory, in Pasadena, Calif., has been exploring, with the CSIRO, the possibility of testing artificial intelligence for drones, rovers and satellites — not yet developed but intended for future space exploration — on the fires. This software would need to withstand extreme conditions on other planets, like “hot temperatures, low visibility and turbulent winds,” said Natasha Stavros, a science system engineer at J.P.L., in an email.
A November 2019 study by J.P.L.’s Blue Sky Thinktank, on which Dr. Stavros was an author, found that the fire-management technologies offering the highest return on investment were autonomous micro-aerial vehicles — small drones typically weighing less than a quarter of a pound — that would be able to navigate themselves through wildfires. Eventually, these drones would operate in autonomous groups or “swarms,” which could monitor wider areas. Their ability to communicate with one another and a distant control center could potentially be used in exploring other planets.
On Earth such drones, equipped with infrared sensors, could also read the heat signatures of plants to determine how stressed the vegetation is in an area — and thus how dry and fire-prone the terrain might be. On the International Space Station, a similar sensor (though not yet small enough to fit on a drone) called Ecostress has been measuring the temperature of plants for almost two years.
As Australia seems to have entered a new era of more extreme and frequent fires, researchers, firefighting organizations and the government increasingly are also looking at ways to help the environment itself adapt in the long run.
Scientists with the University of Melbourne Bushfire Behavior and Management group have developed the Fire Regime Operations Simulation Tool, or FROST, which aims to predict fire behavior over the course of the next century, by taking into account how vegetation transforms after it is burned. Major trials are expected to begin within the next year.
FROST takes uncertainties into account using Bayesian networks, predictive statistical tools that are designed to ask “What if?” of every assumption and then produce a range of possible outcomes in response.
Faced with live fires, firefighters need to decide within a matter of minutes what to defend. Wildlife and vegetation inevitably come second to people and property. By simulating long-term risk, FROST can help find and protect zones for particular wildlife or plant species within a fire-prone area that are less susceptible to the flames.
In late January, Trent Penman, a bush fire risk modeler who leads the group that developed FROST, used the program to identify areas that might act as refuges for a species of tree known as the alpine ash, which is particularly vulnerable to the increasing frequency of wildfires. Alpine ash trees die in high-intensity fires, regenerating from seeds left in the ground. But these seedlings take 20 years to reach maturity. Should the area burn again before then, the young trees will die before any new seeds have been left behind.
Alpine ash is at a tipping point, Dr. Penman said. Extreme fires occurring over the next decade could mean the species becomes endangered “very, very quickly.”
A 2015 paper by academics from the University of Tasmania and the University of Melbourne found that there were 97 percent fewer young, regenerating trees in alpine ash forest sites that had burned twice in 20 years. “Under rapid global warming, which is likely to increase fire frequency, it is hard to be optimistic about the long-term survival of the bioregion’s remaining mature alpine ash forests,” the authors of the paper wrote.
Advancements in technology are important, said Mr. Mullins, the former NSW Fire and Rescue commissioner, but the “big ticket item” is tackling climate change. “It’s a bit like going to a gas fire and putting out all the houses and burning cars around it but not turning off the gas. Well, it’ll keep burning. All the houses, everything: doesn’t matter how much water you put on them, they’ll keep catching fire again.”
“To firefighters it’s pretty simple,” he said. “Deal with the basic problem and all the other problems will go away, eventually.”