Australian authorities are increasingly turning to artificial intelligence (AI) to help predict, detect and prevent fires in Australia's forests.
From teaching AI to detect smoke rising from timber plantations to scanning bushwalkers' photos to assess fuel loads, the technology is becoming an essential part of protecting the nation's forests.
In an Australian first, AI and satellite technology is monitoring the Green Triangle, one of the nation's largest plantation forestry regions, along the South Australian-Victorian border, for bushfires.
Green Triangle Fire Alliance general manager Anthony Walsh said quick fire detection and response was vital in a region where 25 per cent of the land was covered by forestry estate, nestled among farms, townships and natural bushland.
"The sooner we know about fires, the sooner we can put them out, the less damage there is to the entire community," Mr Walsh said.
The Pano AI system has been implemented for the first time commercially in Australia this fire season and is slated for eight fire towers in South Australia and several more in Victoria.
Mr Walsh said the AI system scanned a 20-kilometre radius for signs of smoke.
"We have two cameras on each location. Each view about 180 degrees and those images are stitched together so you have an unimpeded view," he said.
The system also requires a "human in the loop", with the AI at times struggling to differentiate smoke from clouds or dust.
"The last thing we want is just too many notifications. People will stop paying attention," Mr Walsh said.
"It [the feed] goes through a person, and they review it and will say, 'Yes, it is a fire' and then we will be alerted to that, or they'll say, 'No, false alarm'.
"The camera has a 30x optical zoom on it, so they [a human operator] can take over the camera and zoom in and determine what it is."
Timber production under threat
Australian National University professor David Lindenmayer said AI was needed to prevent Australia losing more of its valuable forests.
His team was part of an international study that found up to 35 million hectares of timber-producing forests were lost to fires in the 20 years up to 2021, putting global timber production under significant threat.
It also found Australia was among the worst-affected nations.
"What that means is by proportion, Australia is losing more of its wood-production forest to extreme or high-severity wildfire than any other place on the planet, except Portugal," Professor Lindenmayer said.
"We need to start to embrace new technologies to be able to detect fires much more, and be able to suppress those fires much more quickly.
"Not only drones, not only AI, but also things like lightning-strike modelling, because we know that there are parts of landscapes that are much more susceptible to ignitions from dry lightning.
"But also to use our new insights into the flammability of vegetation to understand where there's ignition places that are likely to start high-severity wildfires."
Citizen science has a role
Assessing forest fuel loads in a nation as vast as Australia is a major challenge, which is where researchers hope the public can combine with AI to play a role.
Researchers at Adelaide University and the University of the Sunshine Coast have developed a mobile phone app that scans user-submitted photos to predict forest fire risk.
Adelaide University professor Javen Shi said AI scanned images, uploaded to the NOBURN app, for bushfire hazards, in the way human experts would do in the field to estimate the severity and possible spread of a potential bushfire.
"Our AI can estimate the fuel load distributed around Australia, so authorities would know which part may have accumulated higher risk, so they may need to do some backburning or other activities," Professor Shi said.
AI could also be used to provide a risk analysis of areas too remote or dangerous for people to explore.
"[The] more people could take photos, that would be better, but we also understand not every park is safe for everyone to go, and that's where smart AI can come in," Professor Shi said.
"Causal AI can see a million problems as one problem. In those parks or forests where you don't have enough data, you learn the pattern from many other different data sets. It's kind of like what ChatGPT was doing."
Professor Shi hopes the app will also spark greater public interest in the potential for AI to help predict and prevent bushfires, as well as better allocating firefighting resources.
"Authorities, experts and citizens — they don't know enough about the potential of AI to help in fighting bushfires," he said.
"This app is just the beginning."