Expert Q&A: Advancing wildfire predictions to protect Canada's forests

By Cultivate Labs on August 28, 2024

In many parts of the world, changing climate conditions are increasing the frequency and intensity of wildfires. This is especially true in Canada where forests cover a substantial portion of the landscape, putting much of the country at risk of wildfires (NRCan).

Forest fire management is an important area of study for the Canadian Forest Service (CFS), an organization within the federal government which conducts science-based research on Canada’s forests. Cultivate Labs is working with researchers at CFS on a crowdsourced forecasting program, “Firesight,” which aims to better anticipate how the changing climate will impact wildfires by capturing predictive trends that can help inform decisions on fire preparedness and mitigation.

Mike Wotton is a Forest Fire Research Scientist at CSF who works closely with fire management agencies across Canada to integrate research models, forecasting, and scenario planning into daily fire management operations. Mike and his team use Firesight to predict wildfire occurrences and damages across Canadian regions. We talked with Mike to learn more about how Firesight is supporting fire management. Here is what he shared:

Q: Why did the Canadian Forest Service decide to initiate a crowdsourced forecasting effort?

A: The wildfire management community has always been in the business of predicting and preventing fires – but the situational and dynamic nature of wildfires poses a variety of challenges for forecasting both in the short term and the longer term. Seasonal forecasting is thought by many in the community to not be possible with any degree of skill, yet many in the community engage in watercooler conversations prior to the wildfire season about what they expect to come in the year.

A few fire management professionals got together to discuss our predictions about the upcoming fire season, and we left determined to explore if harnessing the skill, wisdom, and experience of practitioners across Canada’s fire management community could improve these long term outlooks. With Firesight, we are able to tap into a broader pool of knowledge and perspectives, which is especially valuable given the regionally varied nature of wildfires across Canada. We ultimately hope to improve our wildfire predictive capabilities. 

Q: What are you looking to achieve with Firesight and what value will it bring to the team?

A: Firesight aims to improve the thought process behind decision-making in wildfire management, a field often characterized by uncertainty, dynamic situations, and incomplete information. The primary goal isn't necessarily to produce highly accurate forecasts, but to widely enhance predictive skills within the fire management community and provide a structured environment for testing and measuring regional wildfire forecasts. By asking participants to weigh in on verifiable questions, Firesight creates opportunities for learning, reflection, and comparison of outcomes for our wildfire management community.

Crowd forecasting allows skilled individuals in our fire management community to test and improve their predictive abilities, while also educating our community about trends and conditions impacting wildfires. This process promotes reflection on actual outcomes and comparison with others' predictions. It also seeks to address the lack of clear feedback and evaluation methods for forecasts in this challenging field.

Q: Given the urgency of fire management, how do you approach the time horizons of your forecast questions? What are advantages to short- vs long-term questions, or vice versa?

A: Both have their place and offer different learning opportunities for our community members. Our community is particularly good at very short-term predictions, such as day-to-day fire occurrence. These questions tap into our existing strengths and operational knowledge. We also look at longer-term forecasts for seasonal aspects of fire activity, including metrics like total area burned and the need for international resources. These questions encourage broader strategic thinking and the consideration of higher uncertainty.

We are still finding the right balance to engage participants given a few challenges with forecast horizons. For example, some participants appear skeptical of long-term fire season forecasts, believing them to be inherently unpredictable. Others avoid very short-term questions, feeling these are better suited for fire intelligence staff. This year, we grouped our forecasting “challenge areas” by time frames. For example, there was a challenge for “weekly fire occurrence” predictions and another for “seasonal outlooks”, so we could gauge how effectively different time horizons allowed us to maximize participation, while still generating valuable insights for our wildfire management strategies.

Q: You are in the initial stage of growing and gaining awareness for Firesight across your community. What are the current priorities at this point?

A: We're currently focusing on two main objectives. One is building a foundation; we're working to increase grassroots participation and develop a comprehensive understanding of the potential and limitations of crowd forecasting within our field. Having this solid foundation in place will help us be able to integrate forecasting results more directly into our operations in the future once we’ve established a strong base of participants.

Skill development is another objective. We want to help our team members develop and refine effective forecasting practices. These skills are valuable across a wide range of decision-making scenarios in the seemingly unpredictable world of wildland fire management.

Q: Have there been noticeable learnings for forecasters from this process so far?

A. We are covering a wide geographic range in our forecasting, so this is forcing forecasters out of their regional “comfort zones.” The breadth of perspective that’s needed to forecast creates more awareness of how fire seasons unfold across the country, which strengthens our ability to look out and think about resource availability.

Often during a fire season, it will be really active in one area and less active in another area on the opposite side of the country due to the way big weather systems can organize themselves. In other seasons, like in 2023, we saw a lot of fires happen everywhere at once. So, it’s positive to see forecasters broadening their outlook and expanding their capacity to make predictions beyond their regions. This will be an important skill for the future of fire management.