A study recently published in Scientific Reports, an open access journal from the publishers of Nature, details a first-of-its-kind wildfire simulation model that, in its current iteration, is capable of predicting which buildings in a community are most vulnerable to wildfires and even the amount of damage they will take.

Until now, wildfire prediction models were built with wildlands in mind. Lead study authors Akshat Chulahwat and Hussam Mahmoud, both of Colorado State University, have unveiled the first community-focused wildfire path prediction model. This technology has the potential to be very helpful in optimizing wildfire mitigation by highlighting high-risk buildings that will allow the fire to spread more easily, where buffer zones should be created, and more.

Predicting wildfires with graph theory

The foundation of the model is graph theory, which is a branch of discrete mathematics focused on the study of networks made up of points connected by lines. A common use case of graph theory is predicting how communicable diseases will spread throughout a populace. In a press release about the new study, Mahmoud said that at a broad level, wildfires and diseases behave in similar ways, so it’s possible to apply disease prediction methodology to wildfires and find the most likely paths the wildfire will take from building to building.

After developing the model, Mahmoud and Chulahwat ran initial tests with data from the 2018 Camp Fire and the 2020 Glass Fire provided by wildfire science company Technosylva. The model predicted which buildings were burned or survived with 58 to 64 percent accuracy. According to the press release, the pair have since improved the model by adjusting how it weighs certain factors. After the adjustments, it predicted the Camp Fire’s behavior with 86 percent accuracy.

Mahmoud announced that they will be presenting their work in Sharm El-Shiekh, Egypt at the 2022 United Nations Climate Change Conference.