Effective early detection of forest fires can be achieved by specialised systems of tower-mounted cameras. Foresters and locals with intimate knowledge of the terrain traditionally plan the tower site locations – without the aid of computational optimisation tools. However, such knowledge and expertise may not be available to system planners when entering vast new territories. The process of selecting multiple tower sites from a large number of potential site locations with the aim of maximising system visibility of smoke above a prescribed region is a complex combinatorial optimisation problem. We present two recent applications of novel site-selection frameworks for tower-mounted camera-based wildfire detection systems (CWDS), which have been under development with guidance from experts from the South African developed ForestWatch wildfire detection system. A novel single-site search framework determined alternatives for 13 proposed sites in South Africa's Mpumalanga province, of which 6 alternatives were chosen over the initially proposed sites. The system site selection framework was showcased in determining a four-camera CWDS layout in South Africa's Southern Cape – significantly improving on the detection capability of the layout initially proposed by technical experts.