Using landscape characteristics to predict risk of lion attacks on humans in south-eastern Tanzania

Human-carnivore conflict has profound impacts on carnivore populations and lives of rural communities. In Tanzania, African lions (Panthera leo) have attacked over 1000 people in the last twenty years. We developed a logistic regression model that predicts probability of lion attacks based on landscape characteristics, creating a risk map for two well-studied districts in Tanzania as well as for three neighbouring districts. Results indicate that probability of attack increases with proximity to villages and in areas with a large proportion of open woodland/bushland and crops. Attack risks are also affected by distance from protected areas and by changes in grassland, grassland with crops, wetlands and bare areas. The statistical model also predicted attacks at the ward level (an administrative unit below district) in the two study districts and three additional neighbouring districts. Thus, the technique has potential to identify underlying landscape-related causes of human-wildlife conflict and predict future high-risk areas.