Named after the Greek goddess of war, backed by data from the CSIRO’s Black Mountain fire laboratory and powered by AI, a new weapon lies at the headquarters of the authorities who rule a state ready to burn.
New fire modeling system Athena maps and predicts where bushfires are likely to spread and automatically detects the lives and properties that may be in the path of a fire.
On Sunday, the start of bushfire season, the Athena map flashed with rippling purple pixels as the system simulated more than 85 blazes sweeping through grass and bushland gutted by the hottest September on record.
After a trial run last year, the NSW Rural Fire Service is lighting a fully operational Athena for the first time in preparation for the worst bushfire season since the Black Summer of 2019/20. At times this season, 200 bushfires devastated the state at the same time.
Athena’s ability to quickly assess the severity, behavior and threat to homes during fires is expected to help the NSW Rural Fire Service decide where to deploy trucks, aircraft and personnel on these terrible bushfire days.
The system highlights homes, care homes, schools and critical infrastructure such as vulnerable power lines and automatically ranks which bushfires require critical attention.
“There is nothing like this in Australia,” said Ben Millington, deputy commissioner of the NSW Rural Fire Service, at the state operations center.
“During a recent visit to the US, we worked with the US Forest Service and Cal Fire and showed them a demonstration of this system. They were really impressed – they don’t have anything like that available.”
Athena also tracks fire engines and aircraft while marking water sources such as backyard pools where tankers can replenish their supplies of incendiary bombs.
Weather data, vegetation maps, fire scenes from crews on the ground, infrared images from line scanner aircraft and fire behavior models developed by the CSIRO all feed into Athena. The system will continue to learn and recalibrate its predictive power as fires break out across the state.
The pyrotron: extracting data from the flames
Athena’s fire prediction is partly supported by the CSIRO’s Spark fire behavior model, which uses data from Dr. Andrew Sullivan’s decades of controlled combustion experiments.
Sullivan, who leads the CSIRO’s Bushfire Behavior and Risks team, has conducted fire experiments for decades to test how wind, humidity and vegetation type affect how a burst of flames can turn into a deadly fire.
Many of these took place in the Pyrotron, a 29-meter-long steel tunnel containing a two-ton wind turbine that can accommodate batches of burning bushfire fuel under controlled conditions to extract sterile data from one of nature’s most chaotic forces.
Sullivan’s lab also conducts field experiments. Those studies spawned the fire danger assessment system used until its first overhaul in half a century last year, and made breakthroughs that showed experts had drastically underestimated the speed at which large bushfires spread compared to smaller fires .
The fire lab also developed a rule of thumb that firefighters on the ground could use to roughly calculate the speed of fire spread.
“You can estimate the spread of the fire as a fraction of the open wind speed,” Sullivan said. “In forests this is about 10 percent of the open wind speed, in grass it is 20 percent.”
Despite these findings, field studies of fires are difficult to replicate because scientists cannot control the weather. Fluctuating wind, rain and humidity conditions affect the ability to keep variables in the field constant – a problem that the pyrotron solves.
“It’s about providing a way to study bushfire fuel combustion in a safe and repeatable way to increase the statistical robustness of our results,” Sullivan said.
Using the pyrotron, the fire laboratory has helped determine the different behaviors of fires in eucalyptus forests, wheat fields, paddocks and pine plantations, improving researchers’ ability to simulate the spread of bushfires with mathematical models.
“The fastest fire we had in the pyrotron over a length of five meters lasted 13 seconds, crazy! It was very, very fast and the wind speed was not excessive,” Sullivan said. “It was a bit of a setback because we had been growing the grass for months and it was done in no time. Together, the field work and laboratory work are helping us build the foundational knowledge needed to improve our operating models that people like the RFS use.”
The first time that the fire behavior forecaster is on site
Dr. Mahesh Prakash, working with Sullivan’s team, oversaw the development of the CSIRO’s Spark bushfire simulator over a six-year period.
Previously, the model was only useful for replicating theoretical fires or planning fire hazards to reduce hazards. Now it’s ready to map real-world bushfires and provide a glimpse into the future of flames.
“It has evolved from a research tool to a risk assessment tool to an operational tool. Now it can be used during a real bushfire. This is the first fire season when the authorities will use them furiously,” Prakash said.
To predict what a bushfire could do in 12 hours, Spark considers four key pieces of data: live weather data from the Bureau of Meteorology, the slope of the terrain, the type of vegetation fueling the fire and the timing of ignition.
“It can predict how quickly the bushfire is spreading, where it is going, how high the flame is, what intensity it is and what radiation it is emitting [heat] Impact too,” said Prakash. “Think about Google Earth and the fire history on that map. That’s the kind of visualization we get.”
In January, the South Australian Country Fire Service used Spark to recreate a 45-hectare bushfire in steep terrain in the Adelaide Hills that left two firefighters injured to better analyze the fire’s impact.
“About 90 percent of the behavior they observed in the actual fire was mimicked by Spark,” Prakash said.
After Australia’s driest September and record heat, the senior scientist is worried about the summer ahead.
“I love [the warm weather]“But I’m starting to worry,” he said.
“Spark has the ability to optimize resources where they are needed most. This has the potential to save lives.”
The Examine newsletter explains and analyzes science with a rigorous focus on the evidence. Sign up to receive it every week.