Forest Fire Prediction
In a country with millions of square kilometers of forests, Canadians are acutely aware of the dangers of forest fires. Standing out in the minds of millions are infernos such as the 2016 Fort MacMurray wildfire which saw tens of thousands of people evacuated and approximately $10 billion in damage. 2017 was the third consecutive year of above-average forest fire impacts; 5,600 forest fires burned 3.4 million hectares, causing hundreds of millions of dollars in damage.
While fires are often a natural process in forests, climate projections show that forest fires will likely become more frequent and intense as we enter a period of greater climate uncertainty and above-average temperatures.
Long behind us are the days when the most cutting edge fire-fighting tools were airplanes with bellies full of water. While instrumental in wildfire suppression, typical methods are nowadays being supported by advanced data-driven tools to support decision-making. Synthesizing patterns of human land use, forest dryness and weather, information scientists can contribute to modern firefighting by creating predictive models that interpret these data structures and derive patterns using advanced algorithms.
Douglas Woolford and his research group in the Department of Statistical & Actuarial Sciences at Western University are using modern data science to better understand wildland fires. They are able to use complex data structures to gain an understanding of patterns in key characteristics of wildfire regimes and integrate that knowledge into wildfire management. Using large, complex datasets from across the country, the Woolford Wildland Fire Science Lab develops models to create risk maps which can help predict where fires may occur given current weather conditions. The models will help fire management forces decide how best to allocate resources and personnel.
The Woolford group’s models are already being used. The Ontario Ministry of Natural Resources and Forestry has been using their models to create fire occurrence prediction maps daily through recent fire seasons. The output from such models feeds into systems such as the Aerial Detection Demand Index, which integrates spatial maps of fire occurrence risk, potential fire behaviour and values on the landscape into a single index that identifies priority areas for detection efforts.
Woolford is currently collaborating with other researchers and fire management agencies on projects that will produce fire occurrence prediction models for the Province of Alberta, as well as a national fire occurrence prediction system. Being able to predict when and where fires can occur will give us a much-needed advantage as we see the severity and frequency of forest fires increase in the coming years.