Edmonton researchers create AI robot to predict Alberta wildfires
'We can look ahead for the next where and when,' says lead researcher Mike Flannigan
Scientists at the University of Alberta hope to harness the powers of artificial intelligence to protect Alberta communities from the ravages of wildfires.
A computer model inspired by the human brain, programmed by forest science researchers at the University of Alberta and the University of Oklahoma, is designed to predict extreme fire weather in northern Alberta.
Described as a "self-organizing map," or an SOM, the program relies on raw meteorological data to generate predictions.
Over time, without direction or outside intervention, the program "learns" from this raw data and makes predictions in real time, said Mike Flannigan, co-author of the study and professor at the University of Alberta's department of renewable resources.
Similar to the human mind, the program can be "trained" to find patterns and draw complex conclusions about future forecasts.
The program could help bolster Alberta's early warning system, allowing front-line staff to better deploy resources and brace for possible evacuations.
Unlike other statistical modelling, the U of A program can provide "more robust" predictions because it accounts for complex factors related to wildfires, such as growth rate and intensity, said Flannigan.
This kind of modelling has been used to predict other extreme weather events such as monsoons, but this is the first time it has been applied to wildfires.
- Ex Machina and getting over our fears about artificial intelligence
- Never mind the robots, what about us?
"We're using self-organized maps but it uses a neural network and it is supposed to mimic how we have neurons in our brains and how they pass information along and how they learn," Flannigan said.
"These artificial intelligence methods do exactly the same sort of thing."
Alberta has, on average, 1,500 wildfires every year, but last season was the most devastating.
Last spring, the Fort McMurray wildfire forced the evacuation of more than 90,000 people and caused $3.77 billion in insurable losses, making it the costliest insured natural disaster in Canadian history.
Like most extreme fire disasters, last year's wildfire was fuelled by battering winds and bone-dry weather conditions, said Flannigan.
Wildfires burn an average of two million hectares per year in Canada, but most of the destruction can be attributed to only a few days of severe fire weather, said Flannigan.
- Fort McMurray wildfire costliest insured disaster in Canadian history
- 'Almost biblical': Fort McMurray wildfire named biggest weather event of 2016
These "spread days" are often associated with large-scale weather systems. Better forecasting could be the first line of defence during the next disaster, he said.
"There are a lot of fires on the landscape, but you need three ingredients for fire," Flannigan said.
"You need the fuel; the trees, the shrubs, the grass. You need the ignition — people and lightning — and you need conducive weather, hot, dry windy weather.
If we can get a better handle on those extremes, then we can be better prepared- Mike Flannigan, researcher
"If we can get a better handle on those extremes, then we can be better prepared … We can look ahead for the next where and when."
After three years of research and testing in "diagnostic mode, the program is ready for the field, but Flannigan said it will take time to get it to the front lines.
"It will take a couple years of hard work on both sides. I'm quite prepared to do that and I'm hoping that Alberta and other organizations across the province are interested as well."
The program was described in a new study titled Automated Prediction of Extreme Fire Weather from Synoptic Patterns in Northern Alberta, Canada. It was published Tuesday in the Canadian Journal of Forest Research.