Previous reviews of wildfires where a fatal firefighter burnover occurred have found that the incidents usually share similar characteristics in terms of the fire environment, such as steep slopes and complex topography (e.g. box canyons). Despite these similarities, systematic identification and communication of the locations where these conditions prevail are rare. In this study we used a presence-only machine-learning algorithm (Maximum Entropy, MaxEnt) coupled with spatial location information from past fatal firefighter burnovers to identify and characterise the environmental variables that are likely to produce conditions suitable for a fatal burnover. Southern California was chosen to conduct the analysis as it has a well-documented history of past fatal firefighter burnovers and a complex fire environment. Steep, south-west-oriented slopes located in canyons with a shrub fuel type were found to be the most dangerous locations for firefighters. The relative danger to firefighters from a fatal burnover is described and summarised at both the 30-m pixel and local watershed scale.