UBC Okanagan researchers analyzed nearly 500 records and found that a lack of reliable data keeps many wildfire studies on the shelf rather than in the field.
Canada's 2023 wildfire season burned more than 18.5 million hectares, or an area larger than Greece.
As fires like these grow more frequent and unpredictable, a team of UBC Okanagan researchers has taken stock of two decades of work applying operations research to the problem and found a field that's matured quickly but still struggles to get its best ideas out of the journals and into the field.
The review, published in Operations Research Forum , was led by undergraduate students Paria Rostamian and Kibele Sebnem Yildirim alongside Dr. Amin Ahmadi Digehsara and Dr. Amir Ardestani-Jaafari, all with UBC Okanagan's Faculty of Management .
"Wildfires are only getting more frequent and severe. Reactive, experience-based decision-making can't keep up with how complex and unpredictable they've become," Rostamian says. "Operations research gives fire managers a way to model that complexity directly, so decisions about where to send crews, when to call an evacuation and how to plan recovery are grounded in data rather than guesswork."
The team screened 464 records down to 177 peer-reviewed studies published between 2000 and 2024, then sorted them by how they use mathematical optimization and simulation to support decisions before, during and after a fire.
Three main approaches emerged:
- simulation models that test how a fire might spread and how suppression crews might respond;
- stochastic and robust optimization, which plans for a range of possible fire and weather scenarios, including worst-case ones;
- decomposition and metaheuristic methods, built to handle problems too large and complex for standard solvers.
The review also identifies where the field is falling short. Data availability remains a major constraint: LiDAR mapping is expensive, satellite coverage is inconsistent, and real-time data from active fire zones is often unreliable. And even where the models are strong, most stay theoretical.
"Operations research has become a genuinely powerful foundation for wildfire management, but it's still underused in practice," says Dr. Ahmadi Digehsara. "A lot of these models are built and published as prototypes, not tools agencies can deploy. Closing that gap between analytical rigour and what's usable on the ground during an active fire is where this field needs to go next."
The researchers point to a few directions for closing that gap, including distributionally robust optimization-which blends probability-based and worst-case planning-and hybrid models that pair operations research with artificial intelligence and machine learning to adapt in real time as conditions change.
For governments and fire agencies, the review points to a few practical starting points rather than purely academic ones.
The authors call for investment in shared, standardized wildfire data, the kind of open, benchmarked datasets that would let agencies compare and validate models against the same conditions instead of rebuilding each time.
They also recommend closer, ongoing collaboration between researchers and agencies to test models against real operational constraints before deployment. They also call for a deliberate focus on interpretable, transparent systems. This would help the people making split-second calls on a fire line to understand and trust what a model is telling them.
"The tools to make wildfire response more evidence-based already exist. What's missing is the bridge between the research and the fire line," says Rostamian. "That's how research becomes more than an academic exercise and starts helping the communities living with this risk every summer."
