Spatio-temporal Assessment of Wildfire Risks to Critical Infrastructure: Predicting Wildfires with Machine Learning

As the climate changes, wildfires pose a growing threat to infrastructure. Wildfires can adversely impact reliability across a wide range of infrastructure sectors: the heat can directly damage electricity distribution network and communication equipment; smoke can disrupt transport and solar power production; and ash can contaminate water supplies. While some regions of the world have extensive experience with wildfire driven infrastructure outages, changes in climate and land use mean that infrastructure owners around the world are now facing these challenges, including those in Great Britain. Therefore, it is essential to understand the impacts that wildfires could have on infrastructure owners’ ability to provide essential services to society.  

We present a methodology to use landcover, vegetation properties, weather, and topography to inform the behaviour and likelihood of wildfires in proximity to critical infrastructure. Simulations across 249 distinct scenarios for Great Britain allowed us to examine the expected behaviour of wildfires, and how this behaviour may change seasonally, under different fuel management scenarios and under extreme heatwave events. This culminated in 316,479-point simulations of fire behaviour. Accounting for local landcover, windspeed, and topography have enabled us to spatially map these scenarios to critical infrastructure assets (power and transportation), enabling visualisation of the changes and impact. Finally, we used a machine-learning based methodology (using landcover, vegetation properties, weather, and topography) to inform the likelihood of wildfires occurring in proximity to critical infrastructure. Simulation using historical recorded wildfire incident data enables model validation and provides insight for climate adaptation planning and resilience enhancement strategies.