When a wildfire starts, it can take less than an hour to grow into an out-of-control burn. That's not much time for fire crews to get to the scene, especially in remote areas.
But what if we could predict the likelihood of wildfires before they start? We could get boots on the ground faster or potentially avoid the danger altogether with a controlled burn or other prevention measures.
Wildfire expert Mike Flannigan is working on a system that predicts the extreme weather conditions associated with wildfires, and he's using artificial intelligence to do it.
Flannigan, a meteorologist, professor in the U of A Department of Renewable Resources and director of Canada Wildfire is working with Ryan Lagerquist, '14 BSc(Hons), a PhD student at the University of Oklahoma's School of Meteorology. They've created a computer program that can sort through historical meteorological data associated with high-intensity wildfires, then use it to predict where extreme weather is most likely to create the right - or, more accurately, the wrong - conditions.
Fire weather is traditionally forecast using precipitation, temperature, winds and relative humidity, but this new algorithm incorporates pressure fields to more accurately predict fire-friendly weather. For example, high pressure systems are associated with warm, dry weather and low pressure systems are indicative of cooler, wet conditions.
- An average of 8,000 wildfires burn about 2.5 million hectares a year in Canada (about half the size of Nova Scotia).
- A few large fires are responsible for most of the area burned, Flannigan says. Fires of more than 200 hectares, only three per cent of the total number of wildfires, are responsible for 97 per cent of the area burned.
The computer program uses a form of machine learning called an artificial neural network, or ANN, which is inspired by the way human brains detect patterns and relationships in data through trial and error. Researchers feed the system data and it makes predictions based on probability. Then researchers tell the system whether it has made the right decision. If it's wrong, the system modifies its approach and tries again.
The ANN produces a self-organizing map. While self-organizing maps are already used to predict weather, this study is the first time researchers have used a self-organizing map in fire management.
It's an example of something AI can do much better and faster than humans, both because of the huge amounts of data involved and the ability to find meaningful patterns in the data.
Looking closely at all available data, including pressure levels, gives researchers a fuller picture. "It's more accurate than using traditional precipitation patterns alone," Flannigan says. "Artificial intelligence can help us better predict wildfire."
The program will build on Canada's current wildfire prevention system, says Flannigan, adding he hopes it will be ready to use within five years. By more accurately predicting weather further out, AI could help protect the lives of Canadians and save communities millions of dollars.
We at New Trail welcome your comments. Robust debate and criticism are encouraged, provided it is respectful. We reserve the right to reject comments, images or links that attack ethnicity, nationality, religion, gender or sexual orientation; that include offensive language, threats, spam; are fraudulent or defamatory; infringe on copyright or trademarks; and that just generally aren’t very nice. Discussion is monitored and violation of these guidelines will result in comments being disabled.