Negative interactions between people and large carnivores are common and will probably increase as the human population and livestock production continue to expand. Livestock predation by wild carnivores can significantly affect the livelihoods of farmers, resulting in retaliatory killings and subsequent conflicts between local communities and conservationists. A better understanding of livestock predation patterns could help guide measures to improve both human relationships and coexistence with carnivores. Environmental variables can influence the intensity of livestock predation, are relatively easy to monitor, and could potentially provide a useful predictive framework for targeting mitigation. We chose lion predation of livestock as a model to test whether variations in environmental conditions trigger changes in predation. Analysing 6 years of incident reports for Pandamatenga village in Botswana, an area of high human–lion conflict, we used generalized linear models to show that significantly more attacks coincided with lower moonlight levels and temperatures, and attack severity increased significantly with extreme minimum temperatures. Furthermore, we found a delayed effect of rainfall: lower rainfall was followed by a significantly increased severity of attacks in the following month. Our results suggest that preventative measures, such as introducing deterrents or changing livestock management, could be implemented adaptively based on environmental conditions. This could be a starting point for investigating similar effects in other large carnivores, to reduce livestock attacks and work towards wider human–wildlife coexistence.